The design of this study was a 12-week, double-blind, placebo-controlled and randomized clinical trial conducted between July 2018 and April 2019 at Kyung Hee University Hospital at Gangdong, Seoul, Korea. Potential subjects were screened for eligibility during visit 1 (week −3 to −1). Non-diabetic adults aged 21–70 years were enrolled in this study. The exclusion criteria were as follows: 1) type 2 diabetes based on the criteria of American Diabetes Association [17]; 2) high aspartate and alanine aminotransferases levels (aspartate aminotransferase [AST] or alanine aminotransferase [ALT] over 3 times the upper limit of normal); 3) serum creatinine levels over 1.5 mg/dL; 4) individuals who had taken medication for obesity, anti-diabetic medications, lipid-lowering drugs or corticosteroids within the preceding 3 months; 5) individuals diagnosed with hyperthyroidism or hypothyroidism, except for subclinical hypothyroidism with thyroid-stimulating hormone < 10 mIU/L; 6) subjects with serious illnesses (e.g., heart disease, psychiatric disease, and cancer); and 7) drug abusers. Finally, a total of 74 subjects were included in the full analysis set, and 75 were included in the safety set (Fig. 1).

This study was conducted according to the guidelines proposed by Kyung Hee University Hospital at Gangdong, and all procedures involving human subjects were approved by the Institutional Review Board (IRB) at Kyung Hee University Hospital at Gangdong (IRB number; 2017-06-030). All performed procedures were in accordance with the ethical standards of the 1964 Helsinki Declaration and its later amendments. Written informed consent was obtained from all study participants before inclusion. This trial has been registered with the Clinical Research Information Service (CRIS), Korea (KCT0004672).

Eligible participants were randomly assigned to the placebo group (n = 38), or the D. morbifera group (n = 36) at visit 2. Next, a simple randomization procedure was performed without stratification using a random number list. All of the enrolled subjects were educated about nutrition and exercise prior to starting the trial. They maintained a balanced diet and regular physical activity. We provided subjects with dietary information, such as the components of well-controlled diet, the importance of food choices, and information how to cook low-fat meals. Efficacy and safety assessments were performed at baseline, and at weeks 4 and 12. All of the participants were instructed to take 2 pills once daily with water 30 minutes prior to eating breakfast.

Body weight was measured to the nearest 0.05 kg and height was measured to the nearest 0.1 cm using a measuring system (GL-310; G-Tech International, Uijeongbu, Korea), respectively. Body mass index (BMI) was calculated as weight in kilograms divided by the square of height. The waist circumference (WC) was measured midway between the lower rib margin and the iliac crest by the same researcher. Fat mass was evaluated using the bioimpedance method (InBody720; Biospace, Seoul, Korea).

Blood samples were collected after fasting for more than 8 hours at baseline, and after 4 and 12 weeks of treatment. The kinetic ultraviolet (UV) method was used to measure AST, ALT and blood urea nitrogen (AU5800; Beckman Coulter, Inc., Brea, CA, USA). Creatinine was measured using a kinetic colorimetric assay (AU5800). Lipid profiles (total cholesterol, TG, HDL-C, and LDL-C) were measured using an enzymatic colorimetric assay (AU5800). Fasting plasma glucose (FPG) was measured using an enzymatic UV method (AU5800). Fasting insulin was measured using a chemiluminescent microparticle immunoassay (Architect i1000 System; Abbott Laboratories, Abbott Park, IL, USA). The hemoglobin A1c (HbA1c) was measured using high-performance liquid chromatography (Variant II TURBO; Bio-Rad, Kyoto, Japan). Serum glycated albumin (GA) was measured at week 4 using an enzymatic method that included an albumin-specific proteinase, ketoamine oxidase and albumin assay reagent (LUCICA GA-L; Asahi Kasei Pharma Co., Tokyo, Japan), and a Hitachi 7699 P module autoanalyzer (Hitachi Instruments Service, Tokyo, Japan).

D. morbifera was manufactured and provided by J-creation Co, Ltd. on Jeju Island in the Korea. The leaves from the plant samples were washed 3 times using 100°C water. The leaves were packed in 5 kg units in burlap. They were then extracted for 15 hours in purified water at 85°C. After removing impurities using a 0.5 μm filter, the extracts were transferred to a storage tank and concentrated 10 times. The 90% of the extracts and 10% of dextrin were blended properly, stirred at 90°C, and freeze-dried at −75°C for 40 hours until the proportion of water contents decreased to < 8%. Finally, the dried extract was ground to 50 mesh powder using a pin mil.

The concentration of D. morbifera extract that lowered body weight or glucose level in the animal study was 200 mg/kg body weight for rats (0.08 mg/cm² of body surface areas). Because the ratio of the body surface area in the tested species (rats) to that of humans is 5.45, 480–960 mg was calculated as the equivalent dose for a 60 kg human being. We decided to administer a medium dose of 700 mg D. morbifera, and one pill contains 350 mg D. morbifera leaf extracts to facilitate dosing in capsule form. All subjects were instructed to take 2 pills once daily within 30 minutes before breakfast for 12 weeks. The compositions of placebo and D. morbifera capsule are presented in Supplementary Table 1.

Diabetes was defined as FPG level ≥ 126 mg/dL, or HbA1c ≥ 6.5% [1718]. Prediabetes was defined as 100 ≤ FPG level < 126 mg/dL, or 5.7 ≤ HbA1c < 6.5% [19]. The MetS was diagnosed in any subjects who met 3 or more of the following criteria: 1) abdominal obesity (the Asia Pacific criteria for obesity based on WC: men ≥ 90 cm and women ≥ 85 cm); 2) TG ≥ 150 mg/dL; 3) FPG ≥ 100 mg/dL or the use of antidiabetic medications; 4) HDL-C < 40mg/dL men or < 50 mg/dL in women, or use of lipid-lowering medications; 5) hypertension based on a BP ≥ 130/85 mmHg or the use of antihypertensive medications. These criteria were adopted from the modified Third National Cholesterol Education Program Expert Panel on Detection, Evaluation, and Treatment of High Blood Cholesterol in Adults and the specific values for WC provided by the World Health Organization and the Korean Society for the Study of Obesity [202122]. The homeostatic model assessment for insulin resistance (HOMA-IR) was calculated according to the following formula: fasting insulin (μU/L) × FPG (mg/dL)/405. The homeostatic model assessment of β-cell function (HOMA-β) was calculated using the following formula [23]: 360 × fasting insulin (μIU/L)/FPG (mg/dL) − 63. Difference was calculated by subtracting the baseline values from follow-up values. The percent change was calculated by dividing the difference by the baseline value and multiplying by 100 (%).

The efficacy endpoints were as follows: changes in glycemic parameters (such as FPG, HbA1c, GA, HOMA-IR, and HOMA-β), changes in lipid parameters (such as total cholesterol, TG, LDL-C, and HDL-C), changes in obesity parameters (such as body weight, BMI, WC, total fat, abdominal fat amount, and changes in systolic/diastolic BP after 4 and 12 weeks of treatment. In addition, the proportion of subjects with MetS and the numbers of MetS components were investigated after 12 weeks.

All of the subjects completed a 24-hour recall dietary record before and at the end of the study. Dietary records that were kept during the study were used to reinforce dietary advice and evaluate the amount of total energy and nutrients intake. Nutrient analysis was quantified using a computer-aided nutritional analysis program (CAN pro 4.0; Korean Nutrition Society, Seoul, Korea).

Safety was assessed by monitoring adverse events (AEs), laboratory results (AST, ALT, serum creatinine, leukocyte count, hemoglobin, and platelet count), electrocardiograms, vital signs, and physical examinations. AEs were defined as those that occurred, worsened, or became serious during the study period. The AEs included all unintended consequences that the individual receiving treatment experienced, regardless of its causality. AEs were categorized as definitely related, probably related, possibly related, probably not related, definitely not related, and with an unknown relationship to the study drug.

The data were analyzed using the SPSS software (version 22.0; IBM Corp., Chicago, IL, USA). Continuous variables which were normally distributed were presented as means ± standard deviations, or as medians (25th–75th percentile) if they were not normally distributed. Either the Student's t-test or Mann-Whitney U test was used to compare the differences between the 2 groups. The 2-way repeated measures analysis of variance (ANOVA) for parametric variables and aligned ranks transformation ANOVA for non-parametric variables were used to compare the groups for changes in metabolic parameters from baseline to each assessment time. A paired t-test or Wilcoxon signed-rank test was used to compare the pre- and post-treatment data within a group. Categorical variables were expressed as numbers and percentages, and were analyzed using the χ² test. P < 0.05 was considered statistically significant.

The subjects' baseline demographic and clinical characteristics are described in Table 1. The mean age was 44.8 ± 12.3 years. The proportion of female participants was 69.0%. The mean BMI was 23.8 ± 3.5 kg/m². The majority of subjects were prediabetic patients (62.2%) with a mean FPG of 103.4 ± 10.8 mg/dL, and mean HbA1c of 5.5 ± 0.3%. There were no significant differences in any baseline variables, including age, sex proportion, body composition, BP, glycemic parameters, and lipid profiles, between the placebo and D. morbifera groups.

After 4 weeks of treatment, there were no differences in the glycemic parameters including FPG, GA, HOMA-β, and HOMA-IR between the 2 groups. The fasting insulin level was the only parameter that significantly decreased compared to baseline only in the D. morbifera group (6.1 [4.5–10.9] μIU/L vs. 5.8 [3.9–7.9] μIU/L, P = 0.043; Table 2). After 12 weeks of treatment, the HbA1c level showed a significant time and group interaction (Supplementary Fig. 1), and the HbA1c level significantly decreased in the D. morbifera group, compared to that of the placebo group (% of change between baseline and week 12: −2.27 ± 3.63% vs. 0.10 ± 5.10%, P = 0.025; Fig. 2). HbA1c at week 12 was also significantly decreased from baseline within the D. morbifera group (P = 0.001; Supplementary Fig. 1). Notably, the HbA1c lowering effect was only evident in the subjects with prediabetes using D. morbifera (Fig. 2), when we classified the subjects into prediabetes and normal groups. The fasting serum insulin level, HOMA-β and HOMA-IR at week 12 were all significantly lower than those of baseline within the D. morbifera group. However, FPG, GA, and fasting C-peptide levels did not differ between or within the 2 groups (Table 2).

There was no time and group interaction until 12 weeks of treatment. At week 4, HDL-C was significantly increased and TG was significantly decreased in D. morbifera group than in placebo group (Table 3). Neither difference within group from baseline to week 12 nor difference between placebo and D. morbifera groups was observed in total cholesterol, TG, HDL-C and LDL-C levels after 12 weeks.

Although no significant differences over time in all indices of body composition were found between the 2 groups, the percentage of abdominal fat at week 4 was significantly increased from baseline and remained static at week 12 in the placebo group, while there was no change from baseline in abdominal fat proportion in the D. morbifera group (Table 4). The daily caloric intake, body weight, BMI, and body fat percent did not differ before and after treatment within each group or between groups (Table 4).

There was a significant difference in systemic BP pattern over time between 2 groups (Supplementary Fig. 2). Systolic BP showed a significant decreasing trend until week 12 in the D. morbifera group, whereas it increased in the placebo group. However, both systolic and diastolic BP at week 4 did not differ between and within the 2 groups.

The systolic BP in the D. morbifera group was significantly lower after 12 weeks of treatment than it was at baseline or in the placebo group (difference between baseline and week 12: −3.94 ± 9.81% vs. 3.34 ± 11.70%, P = 0.005; % of change between baseline and week 12: −2.82 ± 7.68% vs. 3.29 ± 10.15%, P = 0.005; Table 5).

The prevalence of MetS of D. morbifera group at week 12 was significantly lower than that of placebo group (13.9% vs. 36.8%, P = 0.022; Fig. 3), although baseline prevalence of MetS did not differ. Among the subjects without MetS at baseline, 4.2% (n = 1) of the subjects in D. morbifera group (n = 24) developed incident MetS after 12 weeks of treatment compared to 16.7% (n = 4) of the subjects in placebo group (n = 24), which is significantly lower (P = 0.027; Fig. 3).

The numbers of baseline MetS components were distributed similarly between the 2 groups (Supplementary Fig. 3). After 12 weeks of treatment, the number of MetS components tended to decrease in the D. morbifera group, while it increased in the placebo group (Supplementary Fig. 3). The numbers of MetS components at week 12 were significantly lower in the D. morbifera group (1.2 ± 1.2 vs. 2.0 ± 1.3, P = 0.010) than it was in the placebo group.

The proportion of total AEs (5.3 vs. 8.1%, P = 0.627) did not differ between the placebo and D. morbifera groups (Supplementary Table 2). There were no serious AEs in either group. The serum AST level of the D. morbifera group was significantly lower after treatment than it was at baseline. However, the AST level remained within the normal reference range. Otherwise, there were no changes in the laboratory results between the groups before and after treatment.