The initial search was conducted on 10 August 2022 using the electronic databases PubMed, Excerpta Medica Database (Embase), Cochrane Central Register of Controlled Trials (CENTRAL), Allied and Complementary Medicine Database (AMED), and Cumulative Index of Nursing and Allied Health Literature (CINAHL). The second search was performed on 15 December 2023. The search terms used were “obesity,” “EATM,” “natural products,” “ephedra,” and “RCT” (Supplementary Table S1). These three categories were combined using the AND Boolean operator with keywords appropriate for each database. In addition, clinical practice guidelines and literature reviews on herbal treatments for obesity were examined. Only articles that met the inclusion criteria were selected for review.

Two reviewers, HC and JO, independently assessed the risk of bias (ROB) according to the RoB 2 tool published by the Cochrane Collaboration (Jac et al., 2019). The tool assesses the risk of bias in five areas: randomization process, deviations from the intended interventions, missing outcome data, measurement of the outcome, and selection of the reported results. Each of the five areas and the risk of bias were rated as low, of some concern, or high. In accordance with the methodology for systematic reviews of interventions (Higgins et al., 2019), if the two reviewers could not reach an agreement on the assessment of ROB, the disagreement was discussed with a third independent reviewer (JL) to reach a consensus.

Statistical analyses were performed using RevMan 5.4 software (Cochrane Training, London, United Kingdom) and R software (version 4.2.1; R Foundation for Statistical Computing, Vienna, Austria).

Seven of the 16 studies analyzed were conducted in the Republic of Korea, 5 in the United States, 2 in Japan, and 2 in China, with 1,460 participants (748 in the treatment group and 712 in the control group). Cheon et al. (2020) and Hioki et al. (2004) studied only women, whereas the others neither restricted sex nor reported sex-based criteria. The mean age of the patients in each study ranged from 30.8 to 60 years.

Regarding the intervention of the treatment group, only Kim et al. (2008) used a ephedra as a single herb as the intervention. At the same time, the remaining studies used a combination of ephedra and other ingredients and herbs in forms such as EATM prescriptions or dietary supplements. In addition, all studies used a placebo as a control, except for Hackman et al. [50], who used a multi-nutrient supplement without ephedra or caffeine as a control. Azushima et al. (2015) used conventional therapy for hypertension in both groups, whereas Cheon et al. (2020), Hackman et al. (2006), Coffey et al. (2004), and Feng-Hao et al. (2012) did not use or mention combination therapy. The remaining studies used the usual care for obesity in both the treatment and control groups. The group that received 5% of the treatment group’s dosage was considered the placebo group.

The treatment period varied from 4 to 24 weeks, and the main outcome measures included BMI, BW, and WC, which were also the selection criteria for this study. Additionally, various metabolic and drug toxicity markers were assessed, including serum lipids, blood glucose, aspartate transaminase (AST), alanine transaminase (ALT), and blood urea nitrogen (BUN). The daily dose of ephedra varied from 0.6 to 14 g. The details of each study are summarized in Table 1, and the composition and dosage of the interventions are summarized in Supplementary Table S2.

None of the analyzed studies demonstrated a low overall risk of bias, whereas all studies had some concerns or a high overall risk of bias. The risk of bias for the outcome measurement was low because the outcome measure in all studies was an objective metric. Except for two studies (Boozer et al., 2001; Boozer et al., 2002) that reported measurements of height and weight without reporting BMI as an outcome, all other studies reported all outcomes described in the Methods. Therefore, the risk of bias in the selection of reported results was mostly low. However, in some studies (C et al., 2004; Azushima et al., 2015; Feng-Hao et al., 2012; MA et al., 2014), a third party did not perform the randomization process or was not properly blinded, raising concerns about the potential for bias in the randomization process. In addition, the risk of bias was high for deviations from the intended interventions and missing outcome data owing to the occurrence of adverse effects that could be inferred from the intervention, and a high number of dropouts for unclear reasons (Figure 2).

A meta-analysis was conducted to examine the effectiveness of EOMs in reducing BMI, BW, and WC. First, 11 studies reported the change in the BMI as an outcome measure, of which 5 reported the amount of change and SD value, while 6 reported the final value and SD value. Forest plots were generated to present each subgroup’s changes and final values, and the overall effects were pooled. The study by Hackman et al. (2006) was excluded because the SD values and raw data were not available. The meta-analysis revealed that EOMs resulted in a statistically significant additional reduction in the BMI compared to the control group (MD = −0.38 kg/m²; 95% CI= −0.68 to −0.09) (Figure 3A).

In addition, 13 studies reported changes in BW as an outcome measure, of which 8 reported changes and SD values, while 5 reported final values and SD values. The study by Hackman et al. (2006) was excluded for the same reason. The meta-analysis showed that EOMs resulted in a statistically significant additional reduction in BW compared to the control group (MD = −1.5 kg; 95% CI= −2.46 to −0.54) (Figure 3B).

Nine studies reported changes in WC as an outcome measure, of which six reported change and SD values and three reported final and SD values. The meta-analysis presented that EOMs resulted in a statistically significant additional reduction in WC compared to the control group (MD = −1.13 cm; 95% CI= −2.05 to −0.20) (Figure 3C).

The studies by Kim et al. (2008), Coffey et al. (2004), Boozer et al. (2002), and Feng-Hao et al. (2012), which did not report the number of adverse events in the treatment groups, were excluded, the remaining 12 studies were examined. Adverse events were classified into several categories: cardiac, gastrointestinal, serum hepatic enzyme levels, headache/neuropsychiatric, autonomic hyperactivity, gynecological, and dermatological symptoms. The types and numbers of adverse events are summarized in Table 2. The number of dropouts due to these adverse effects was also included. The most frequently reported adverse events were headache and neuropsychiatric symptoms, with headache being the most common. Additionally, the number of adverse events, participants who experienced adverse events, dropouts, and reasons for dropouts in both the treatment and control groups for all 16 studies are summarized in Table 3.

A meta-analysis was performed using data from five studies that reported the number of participants who experienced adverse events. There was no statistically significant difference between the treatment and control groups (RR = 0.99; 95% CI = 0.80–1.21) (Figure 4A). A meta-analysis of the number of dropouts across all selected studies showed no statistically significant differences between the treatment and control groups (RR = 0.99; 95% CI = 0.83–1.17) (Figure 4B).

In the meta-analysis of post-treatment changes in the BMI, BW, and WC, the chi² test yielded a value of p < 0.10, and the I² values were 62, 87, and 50%, respectively, indicating high heterogeneity among the studies. Therefore, a univariate meta-regression analysis explored the relationship between heterogeneity and covariates. For change in BMI, both the daily dose of ephedrine (QM = 18.5143, p < 0.001, tau² = 0, I² = 0%, and R² = 100%) and duration (QM = 4.4590, p = 0.0347, tau² = 0.0313, I² = 29.34%, and R² = 69.50%) explained the heterogeneity but not the initial BMI (QM = 1.0719, p = 0.3005, tau² = 0.0625, I² = 44.22%, and R² = 39.20%). For changes in BW, the daily dose of ephedrine (QM = 14.0831, p = 0.0002, tau² = 0.3087, I² = 37.74%, and R² = 85.93%) explained the heterogeneity, but the period (QM = 2.3635, p = 0.1242, tau² = 1.5880, I² = 83.14%, and R² = 27.62%) and initial BW (QM = 3.4327, p = 0.0639, tau² = 1.4732, I² = 81.47%, and R² = 32.85%) did not explain the heterogeneity. Finally, for changes in WC, the period (QM = 9.8250, p = 0.0017, tau² = 0, I² = 0%, and R² = 100%) explained heterogeneity as a variable, but the daily dose of ephedrine (QM = 2.7709, p = 0.0960, tau² = 0.4313, I² = 33.55%, and R² = 48.71%) and initial WC (QM = 0.0012, p = 0.9724, tau² = 1.1167, I² = 55.66%, and R² = 0%) did not explain the heterogeneity (Table 4). The results of the meta-regression analysis are visualized using bubble plots (Supplementary Figure S1).

We also generated bubble plots for 10 studies reporting changes in BMI, 12 studies reporting changes in BW, and 9 studies reporting changes in WC, all of which reported the daily doses of ephedra or ephedrine. We developed a four-dimensional graph with the x-axis indicating the baseline measurement of the outcome variable, the y-axis indicating the daily dose of ephedrine, the color of the bubble indicating the study period, and the size of the bubble representing the reduction in the outcome measure (Figures 5A–C). In the same study, we developed another type of bubble plot for each outcome measure. The x-axis represents the outcome measure at baseline, the y-axis represents the total ephedrine dose, and the bubble size represents a reduction in the outcome measure (Supplementary Figure S2).

We assessed the potential for publication bias in the outcome variables, BMI and BW, using 10 or more studies included in the meta-analysis. Contour-enhanced funnel plots showed asymmetry in all outcome measures, confirming the possibility of publication bias (Figures 6A, B). Egger’s test was used to assess publication bias. There was no statistically significant publication bias for BW (p = 0.3361); however, there was a possibility of publication bias for BMI (p = 0.0300).

This study aimed to determine the efficacy and safety of EOMs in overweight and obese individuals. The 16 RCTs that included ephedra in interventions for individuals with a BMI of 25 or greater were identified, and their characteristics, interventions, and adverse effects were summarized. A meta-analysis was performed on the outcome measures of BMI, BW, and WC to determine the statistical significance of these effects. The results showed a statistically significant reduction in all outcome measures in the treatment group compared with the control group. A meta-analysis was also performed on adverse events and dropouts and showed no statistically significant difference between the treatment and control groups. Furthermore, a meta-regression analysis was conducted to explore the relationship between heterogeneity in the meta-analysis of effects and covariates. The results indicated that the daily dose of ephedrine and treatment period explained the heterogeneity in BMI change, a daily dose of ephedrine explained the heterogeneity in BW change, and the treatment period explained the heterogeneity in WC change. Additionally, bubble plots were used to visually demonstrate that the covariates were proportionally associated with reductions in the outcome measures (BMI, BW, and WC).

This study has several strengths. Although previous reviews have examined herbal prescriptions for obesity, as well as single agents like ephedra and ephedrine (Park et al., 2012; Shekelle et al., 2003; Sui et al., 2012), no review has focused solely on ephedra. Hence, this systematic review is the first to comprehensively examine the use of ephedra, including single agents, mixed herbal formulas, and food supplements, for weight loss and treating obesity. Moreover, the selection of studies from diverse countries, formulations, and treatment periods allowed us to extensively evaluate their characteristics, therapeutic effects, and adverse effects. Notably, we summarized the types and frequencies of adverse events reported in the included studies to help identify possible trends in adverse events.

Additionally, dosage, duration, and patient baseline are important considerations in clinical practice. This study used meta-regression and bubble plots to explore the relationship between these factors and treatment effects. The results of the meta-regression analysis showed that the daily dose of ephedrine and the treatment period explained a significant amount of heterogeneity in the therapeutic effect. The bubble plot also illustrates the potential relationship between these factors and their therapeutic effects. These findings may provide a basis for further research into the dose–response relationship between treatment period and ephedra dose in treating obesity.

However, this study should be interpreted with caution due to the following limitations. First, the majority of the interventions included in the studies encompassed ephedra and other herbs or chemicals that may have contributed to the treatment effect. Therefore, it cannot be concluded that the results of this meta-analysis are solely attributable to ephedra. In order to evaluate ephedra in combination formulations, further pharmacokinetic and pharmacodynamic studies are required, considering drug interactions between herbal ingredients (Luo et al., 2020). Currently, various studies, including network pharmacology analyses and pharmacokinetic experiments, are being conducted (Miao et al., 2020; Song et al., 2015; Wei et al., 2014). It is deemed necessary to carry out additional research on the synergistic effects of ephedra in order to establish safe guidelines for its clinical use. However, it is noteworthy that EOMs have been reviewed and found to be effective. In clinical studies of EATM, it is impractical to administer ephedra as the sole agent. Therefore, to obtain data on human subjects, an alternative approach would be to analyze studies on EOMs.

Among the various active compounds in ephedra, ephedrine is recognized for its significant role in weight reduction (Miao et al., 2020). To explore the dose effect, this study assumed that ephedra formulations contained the same amount of ephedrine, 5.25 mg/g, the minimum content threshold suggested by the Korean Pharmacopoeia. However, ephedra formulations comprise natural, unprocessed products, and their ephedrine content varies depending on the origin and time of harvest (Matsumoto et al., 2015). There are also variations in the chemical composition of different species of ephedra, and little is known about the clinical differences caused by different alkaloids (Ibragic and Sofić, 2015). Therefore, the results of this study cannot be directly extrapolated to clinical practice. Given the variability due to individual differences between species, along with the effects and interactions when combined with other herbs and prescriptions, further research is necessary, as previously mentioned.

During the study selection process, several limitations were identified in the selection of outcome measures. First, we focused solely on whether the outcome measures included BMI, body weight, and waist circumference without imposing additional criteria. Consequently, the studies we included did not uniformly address the same liver function tests and metabolic markers, such as AST, ALT, BUN, leptin, and adiponectin, with some studies omitting these as outcome measures. As a result, we were unable to comprehensively evaluate the effects of ephedra on metabolic markers and liver function. Second, BMI, which considers only weight and height, is a measure that does not account for other factors, such as muscle mass. Specifically, in older populations, BMI is not an accurate predictor of obesity as it may underestimate obesity due to decreased muscle mass and increased body fat with age (Batsis et al., 2016).

Meta-regression analyses are deemed more meaningful when the number of studies is large, and it is recommended that at least 10 studies should be included (Higgins et al., 2019). However, this study was an exploratory attempt with few studies, especially for WC, and only nine studies met the recommended criteria. Therefore, only univariate meta-regression analyses were performed, and there were differences in significant covariates depending on the outcome measure. Furthermore, Egger’s test for publication bias recommends 10 or more studies in which the WC does not meet the criteria. In addition, bubble plots are solely visual graphs and do not offer conclusive evidence suggesting causality. Owing to the limitations mentioned above (variable ephedrine content of ephedra, insufficient number of studies, and limitations of bubble plots), we limited our meta-regression and bubble plots to exploratory attempts at determining dose–response relationships of ephedra. Nevertheless, as mentioned in the strengths section, these preliminary attempts provide a foundation for further research.

Finally, although we selected the daily dose of ephedrine as a key covariate, this does not mean that we considered the weight-loss effects of ephedra solely dependent on ephedrine. As mentioned earlier, ephedra is a multicomponent herb, and many different mechanisms explain its effects, with non-alkaloidal components contributing to its effects. However, several selected studies reported only the ephedrine dose rather than the ephedra dose. Therefore, we chose ephedrine as the baseline metric to compare the ephedra doses across studies.

Adverse effects may be expected to increase with higher doses of ephedra; however, this analysis was limited by the small sample size, which needed to be increased for statistical power. As more high-quality RCTs that include pharmacokinetic studies are conducted, a dose–response analysis of ephedra’s efficacy and adverse effects may suggest a safe and appropriate dose of ephedra using methods such as the restricted cubic spline method (Cj et al., 2022).

However, due to unclear herbal constituents, many RCTs were excluded during the selection process. Even among the selected studies, some did not specify the dose of ephedra. High-quality RCTs that clearly describe the study methods, including constituent herbs and the dose of ephedra, are required to draw useful statistical conclusions in clinical practice. Additionally, there needs to be more uniformity in the presentation of drug ingredients and doses, such as ephedrine or ephedra doses. Thus, an international consensus on labeling standards for intervention composition and dose is needed for RCTs on natural products and herbal medicines.

However, the same drug may elicit different responses. For example, the optimal dosage of warfarin may be influenced by CYP2C9 and VKORC1 genotypes, and leveraging this genetic information in prescribing may mitigate adverse effects (Wang L. et al., 2011). Furthermore, recent studies have attempted to use genetics to predict drug responses, such as detecting drug resistance using gene chips (Yin et al., 2020). In addition, inhibiting and modulating certain genes at the fetal and neonatal stages can lead to a phenotype susceptible to cardiac ischemia (Zhang et al., 2021; Patterson et al., 2010; Lawrence et al., 2011). Integrating these genomic studies with the sensitivity to the effects and adverse reactions of ephedra could eventually lead to the determination of its indications and contraindications through genomic testing.

Exploring the impact of ephedra on different population groups would be valuable as this could offer clinicians a more comprehensive understanding of its use. For example, the effects and pharmacokinetic properties of a drug may vary based on factors such as sex and age (Schwartz, 2003; Davis et al., 2012; Busetto et al., 2009). However, in this study, no restrictions were placed on the population to allow for a more comprehensive analysis. Furthermore, given that our study is a secondary analysis (systematic review) based on a prospective RCT, it does not allow for the detailed assessment of effects across different population groups as we cannot acquire individual patient data. The studies we included aimed to balance the baseline characteristics between treatment and control groups through randomization. Furthermore, our review of existing studies revealed that many had conflicting results or involved small sample sizes, reducing their reliability (Haller and Benowitz, 2000; Samenuk et al., 2002; Gurley et al., 1998). However, as more RCTs incorporate diverse population groups in their designs, or as retrospective studies based on individual patient medical records on ephedra increase, it is anticipated that these avenues will provide broader insights into the effects and use of ephedra.

Metabolic markers such as leptin and adiponectin could serve as alternative indicators of the effect of ephedra on weight reduction or the treatment of metabolic disorders (Klempel and Varady, 2011; Yoon et al., 2011). Conversely, to evaluate safety, several cases have reported that ephedra has the potential to elevate liver enzymes and influence kidney function due to its diuretic effects (Saeed et al., 2019). Therefore, to accurately determine the weight-loss efficacy and safety of ephedra, it is essential to establish guidelines specifying which markers should be included as outcome measures. Incorporating these markers into study designs will contribute to higher-quality research in the future.

In conclusion, this study suggests that EOMs may effectively treat overweight and obesity with higher daily doses or longer treatment periods, possibly resulting in greater efficacy. However, because this may increase the risk of adverse events, clinicians must weigh the tradeoff between effectiveness and adverse effects before prescribing these drugs. We anticipate that further research will contribute to the development of professional and standardized guidelines, ensuring the safe use of ephedra in clinical practice. The necessity to mandate the precise specification of the amounts of ephedra and ephedrine used in clinical research should be thoroughly discussed.