The following food samples were produced by Q'sai Co., Ltd. (Fukuoka, Japan): Placebo; low-dose kale (7 g kale-containing food; active-L); and high-dose kale (14 g kale-containing food; active-H). The ingredients of the food samples are presented in Table I. At the measurement times (test periods I, II and III, in which each experimental procedure was performed in one day), the subjects consumed one of the food samples together with a high-carbohydrate meal (rice with chicken and eggs; 575 kcal energy, 18.7 g protein, 3.2 g fat, 115.6 g carbohydrate, and 1,243 mg sodium in a total meal of 510 g). The meal duration time was set at 10 min. The dose of kale was defined based on our previous clinical trial of its effect on postprandial plasma glucose levels (16). In that randomized, double-blind, placebo-controlled, crossover study, intake of kale-containing food at a dose of 14 g significantly decreased postprandial plasma glucose levels. To investigate the appropriate effective dose, the doses were set at 7 and 14 g in the present study.

A randomized, double-blind, placebo-controlled, crossover study was conducted with the aid of a fund from Q'sai Co., Ltd. at two centers (Kenshokai Fukushima Health Care Center, Osaka, Japan; and Go Clinic, Osaka, Japan) under the supervision of the principal investigator (Dr Sumio Kondo; Kenshokai Fukushima Health Care Center). The study was conducted from April to July 2015. The study conformed to the principles of the Declaration of Helsinki and was approved by the institutional review board of the incorporated medical institution (Aiseikai Aisei Hospital Ueno Clinic, Tokyo, Japan) on the basis of ethical issues, the protocol, and information on the food samples.

The subjects were recruited in April 2015. The details of the study were disclosed to the subjects before the start of the study, and informed consent was obtained from all subjects. The inclusion criteria were as follows: i) Male or female aged between 20 and 64 years; ii) fasting plasma glucose level of ≤125 mg/dl at the pretrial test; and iii) postprandial plasma glucose level of 140–199 mg/dl at 30 min after intake of the high-carbohydrate meal at the pretrial test. The exclusion criteria were as follows: i) Those habitually taking green vegetable juice three times or more per week; ii) those taking health foods or medicines that may affect sugar metabolism; iii) those suffering from any disease or with a case history of a serious disease that required medication; iv) those suffering from serious diseases, such as diabetes mellitus or failure of the liver, kidney, or heart, or with a case history of such diseases; v) those taking warfarin, a vitamin K-dependent anticoagulant that inhibits synthesis of coagulation factors; vi) those who may be liable to allergy in response to kale or the high-carbohydrate meal containing rice, chicken, and egg; vii) those participating in any other clinical trial within 1 month or planning to participate in such trials after providing informed consent to join the present study; viii) those planning to get pregnant or nurse a baby during the study; ix) those judged inappropriate as participants from answers to the lifestyle questionnaire; and x) those judged inappropriate as participants from the view of the principal investigator. The 42 included subjects were 23 men and 19 women aged 44.7±11.1 years. The subjects were assigned to group X (n=14), group Y (n=14), and group Z (n=14) based on random number tables. The allocation was conducted by Yoshihisa Kibune (TTC Co., Ltd., Tokyo, Japan) and concealed from the subjects, the investigators, and the researchers who recruited and assessed the subjects. In the three test periods (periods I, II and III), the subjects in the three groups were assigned to consume the food samples as shown in Fig. 1.

During the trial, the following requests were made to the participants: i) Not to change their lifestyle, including diet, alcohol drinking, and sleep patterns; ii) to avoid over-exercising, abstemious eating, or overeating; iii) not to change their exercise habits; iv) not to take health foods or supplements; v) to log their lifestyle each day; vi) to go to bed by 12:00 a.m. on the day before the tests; vii) to avoid drinking alcohol for 2 days before the test; viii) to avoid consuming any meals after dinner on the day before the test and to continue abstaining from food until the test; ix) not to consume a high-fat meal the day before the test; x) to eat dinner at the usual time; xi) not to drastically change the dietary composition of their dinner on the day before the test; xii) to avoid heavy exercise on the day before the test and the day of the test; and xiii) not to consume any meal other than that provided for the test on the day of the test.

In each of the test periods I, II or III, the participants consumed one of the food samples together with the high-carbohydrate meal (Fig. 1). The plasma glucose and insulin levels were measured before (preprandial) and at 30, 60, 90 and 120 min after the meal. Each participant logged their lifestyle during the 3 days before the test. There were 1-week washout periods between test periods I and II, and between test periods II and III.

The primary outcome measure was plasma glucose level at 30 min following consumption of the high-carbohydrate meal. The secondary outcome measures were as follows: Plasma glucose area under the blood concentration-time curve (AUC) for the plasma glucose level determined using values obtained for up to 2 h following the high-carbohydrate meal (AUC_{0–2 h}); plasma glucose maximum plasma concentration (C_max); fasting plasma glucose level; plasma glucose levels at 60, 90 and 120 min after meal consumption; fasting plasma insulin level; and plasma insulin levels at 30, 60, 90 and 120 min after meal consumption.

The following measurements were performed according to the standard laboratory protocol of LSI Medience Corporation (Tokyo, Japan). Plasma glucose was determined by glucose oxidase-electrochemical assay using a GA08lll automated glucose analyzer (A&T Corporation, Yokohama, Japan). Plasma insulin was determined by chemiluminescence immunoassay using an ARCHITECT i2000 analyzer (Abbott Japan, Tokyo, Japan). The AUC for the plasma glucose level was determined using values obtained for up to 2 h following the high-carbohydrate meal (AUC_{0–2 h}).

Further surveys and testing were performed on the indicated occasions as follows: Lifestyle questionnaire, physical examination and laboratory examination at pretrial testing; a medical interview, and somatometry at pretrial testing and test periods I, II and III.

An artificial digest consisting of 2% (w/v) methylcellulose and 5% (w/v) glucose was added to the powdered kale (1–4.8%; w/v) or kale-derived dietary fiber (0.5–1.2%; w/v), and viscosity of the resulting mixture was measured three times using an FCV-100 ultrasonic viscosity meter (Fuji Ultrasonic Engineering, Hamamatsu, Japan). Values are presented as units of mPa·s.

All of the measured values are expressed as means ± standard deviation. The statistical methods used for analysis were as follows: Analysis of variance (ANOVA) for subject background data; χ² test for subject number; ANOVA for comparisons between groups X, Y and Z and for inter-individual, -period, and -food sample comparisons; Tukey's test for efficacy of the food samples and viscosity results. P<0.05 was considered to indicate a statistically significant difference.

Forty-two subjects were enrolled and randomly allocated to the three groups (groups X, Y and Z). The subjects consumed one of the food samples, which were comprised of a placebo, active-L, or active-H, at each of the three test periods, as depicted in Fig. 1. All of the participants completed the study, but two subjects in group Z violated the directions from the principal investigator. These two subjects were excluded from the analysis of efficacy, but included in the analysis for safety (Fig. 1). The background data for the three groups are presented in Table II. ANOVA revealed no significant inter-group variation in the background data (Table II).

As shown in Table III, no significant inter-period differences in the levels of fasting plasma glucose, AUC_{0–2 h}, or C_max were identified. Furthermore, no significant inter-food sample differences were observed in the levels of fasting plasma glucose. Thus, no obvious bias that hampered analyses of AUC_{0–2 h} or C_max was noted, therefore usual statistical methods for the analysis of efficacy were employed. However, there were significant inter-period differences in the levels of plasma glucose at 30 and 120 min (Table III); therefore, these data were excluded from the statistical analysis of efficacy.

Fig. 2A shows the time courses of the changes in plasma glucose following consumption of the high-carbohydrate test meal. The level increased transiently following consumption of the test meal, with a peak at ~30–60 min, when the subjects consumed the placebo. The increase was significantly smaller when the subjects consumed the kale-containing food (active-L or active-H) compared with the placebo according to Tukey's test (P<0.01 at 60 min). No significant differences were identified between the values obtained with active-L and active-H. The C_max values (mg/dl) were calculated to be 176.0±25.6 for placebo, 163.2±24.2 for active-L (P<0.01 vs. placebo), and 162.0±22.9 for active-H (P<0.01 vs. placebo; Fig. 2B). The AUC_{0–2 h} values (mg/h/dl) were 284.2±43.0 for placebo, 268.4±42.6 for active-L (P<0.05 vs. placebo), and 266.3±41.9 for active-H (P<0.05 vs. placebo; Fig. 2C).

The data are summarized in Table IV. The plasma level of insulin increased transiently after the test meal intake, with a peak at ~30 min, when the subjects consumed the placebo. When the subjects consumed the kale-containing food (active-L or active-H), the increases in plasma insulin level were moderate, and the differences were not significant.

The addition of powdered kale, the active ingredient of the test food, at 1–4.8% (w/v) resulted in a dose-dependent increase in viscosity of the mixture (Fig. 3). Furthermore, kale-derived dietary fiber exerted a potent effect when added at 0.5–1.2% (w/v). The effect exerted by kale-derived dietary fiber at 0.5% (w/v) was comparable to that exerted by kale powder at 4.8% (w/v; Fig. 3).

There were no adverse events considered to be caused by consumption of the kale-containing food, based on the judgment of the principal investigator after examining data for physical measurements, blood biochemistry, hematology and urinalysis, as well as medical interview.