The experimental protocol was approved by the ethical committee from the Hospital General de Zona 46 from the Instituto Mexicano del Seguro Social (IMSS). A total of 30 patients with type 2 diabetes according to the WHO criteria, were recruited from subjects attending the Hospital. The purpose of the study was explained to the participants and they gave written informed consent. Subjects were included if they had less than 10 years of type 2 diabetes diagnosis, were obese according to the WHO criterion (Body Mass Index, BMI > 30), were between 40 to 60 years of age, had maintained stable weight during three months prior to experimentation and were under the attention of a health care provider from the IMSS. Study participation did not limit the medical management they were provided. Subjects not included in this study were BMI < 30, with complications such as nephropathy or hepatic diseases, on psychiatric treatments, receiving inmmunosupressors or body reducers, with a smoking or alcoholic history and receiving insulin treatments. Seventeen subjects were taking oral hypoglycemic agents, 11 were on biguanides alone and six were taking both sulfonylureas and biguanides. A criterion for exclusion was if they modified their medical treatment during the study. All participants were examined for anthropometric measurements including height, body weight, waist and hip circumferences. Other determinations were blood pressure and fat percentage. A 24 h food record was performed by a nutritionist using a validated survey.

A cross over design with two experimental periods of 4 weeks was used to compare the effects of NBS supplementation with those of soya milk. Thirty patients met the eligibility criteria, complete baseline testing and were randomly assigned to two groups of fifteen subjects to receive either 24 g of NBS dissolved in 240 mL of water per day or 24 g/day of soy milk dissolved in the same volume of water for 4 weeks. After this period all data measures were collected, groups cross over to the alternate treatment group for an additional 4 weeks period after which final measures were taken at 8 weeks. There was no wash-out period between the interventions. All participants received both supplements, and on the two phases of the experiment they were instructed not to modify their diet and exercise habits. Two women, one of each group, withdrew during the first phase due to personal reasons. The treatment compliance was supervised every week through clinical interviews in order to check the amount of consumed supplement and determine their body weight, side effects or adverse events. Further information was obtained from the patients through telephone interviews.

The NBS was provided by the Centro de Investigaciones Agropecuarias of the Universidad Juarez Autonoma de Tabasco. NBS was obtained from unripe (green) banana (Musa cavendish AAA) with a physiological age of 15 weeks obtained from a packing plant situated on Km 43.5 of the Villahermosa-Teapa road. The NBS was isolated by a previously described procedure [14]. Briefly, after washing the bananas were peeled, cut into 5–6 cm³ pieces, immediately rinsed in citric acid solution and then macerated at low speed in an industrial blender for 2 min. The homogenate was consecutively sieved through screens (30, 80 and 100 US mesh) and washed with distilled water; it was then centrifuged at 10,000 rpm. Sediment was further purified by washing and centrifugation. The white starch sediment was dried in a convection oven at 40–45 ºC, passed through a 100 mesh screen and stored at room temperature in sealed glass jars. The dry basis yield of starch was 60%. The proximate analysis was as follows: 3.4% of moisture content, 1.88% protein, 0.4281% fat and 0.78% ash (AOAC recommended methods 14.003, 14.057, 14.059 and 14.006). Water activity (a_w) of NBS was measured using the Aqualab (Decagon) equipment and gave a result of 0.59.

The determination of resistance in the banana starch flour was carried out according to the enzymatic method of Goñi et al. [15] as follow: protein residues in the powder were hydrolyzed with pepsin under acidic conditions (pH 1.5, 40 ºC, 1 h). Then, tris-maleate buffer was added and the mixture incubated with α-amylase (37 ºC, pH 6.9, 16 h) to hydrolyze digestible starch. The hydrolysis product was eliminated and the non-digestible starch was recovered by centrifugation. After dispersing under alkaline conditions (2M KOH) to solubilize NBS, it was incubated with amyloglucosidase under acid conditions. Glucose concentration was determined by an enzymatic-colorimetric method. Resistance starch concentration of NBS was measured as percentage on a dry weight basis. A 34% of resistance was found by this method. Twenty four grams of native banana starch contained 8.16 g of resistant starch.

Soy milk powder was purchased as the commercial product Soyapac produced from whole soy beans by Colpac (Km 3, Navojoa-Sonora road, Huatabampo, Navojoa, Sonora, Mexico) with a nutritional content of 46.2% of fat, 30.8% carbohydrates, 20.5% protein.

Blood samples were obtained by trained personal at the onset of the study and then after 4 and 8 weeks. Blood serum samples were immediately frozen and conserved at −70 ºC until biochemical determinations. Sera samples were analyzed for glucose, cholesterol, LDL, HDL-cholesterol and triglycerides by using an ADVIA® 1200 Chemistry System from Siemens Healthcare Diagnostics (intra-assay coefficient of variation: glucose 3.5%, cholesterol 4.9%, LDL 5.2%, HDL-cholesterol 6.1%, triglycerides 5.2%). Insulin was determined according to the method for AxSYM of Abbott based on a microassay microparticle inmunoenzymatic assay. To prevent enzymatic insulin degradation special care was taken to avoid hemolysis during the blood sampling and manipulation.

Glycated hemoglobin (HbA_1c) was measured by a turbidimetric inmunoinhibition method according to Beckman Coulter for a Synchron CX7 Clinical System. Insulin resistance was estimated according to the Homeostasis Model Assessment (HOMA) which was calculated by the product of the fasting concentrations of glucose (mg/dL) and insulin (μU/mL) divided by 405 [16].

The characteristics of the subjects are presented as mean ± SD or median and percentiles. D’Agostino-Pearson normality test was performed to assess if the data were consistent with a Gaussian distribution. Changes of endpoints after treatments at 4 weeks were expressed as change from baseline (1st treatment) and for the 2nd treatment baseline was considered the endpoint of the 1st treatment. Student’s t test was used for comparing variables with normal distribution or Mann-Whitney test for data from a nongaussian distribution. With the purpose to investigate the possible carry-over effects during the second treatment period we carried out the statistical analysis on the first period as a parallel design. Statistical significance was defined as p < 0.05. Calculation was performed using GraphPAD PRIZMA software version 5.01.

The baseline characteristics of 28 patients who completed the treatments are shown on Table 1. A great percentage of the obese subjects were women (85.7%), most of them housewives, 96% of them had waist circumference values over 88 cm, with a high waist-to-hip ratios, 0.873 (0.85, 0.91; median and 25th,75th percentiles). Half of the male subjects had WHR values higher than 0.95. Poor glycemic control was observed in most patients, despite their medical management, 85% had glucose values over 126 mg/dL and 42% had HbA_1c percentage higher than 7.0. Cholesterol levels were higher than 200 mg/dL in 60.7% of the participants and 76% had triglycerides over 160 mg/dL. In addition, 94% of patients had a sedentary lifestyle and most of them lived in rural areas and had both a low academic and socioeconomic status.

Compliance with the diet and treatment was excellent (91% for NBS and 93% for control group). No significant differences in compliance were observed between the supplemented groups. The NBS preparation was in general well tolerated, although three patients from each group reported flatulence, increased number of evacuations and soft feces during the first week of the experimental period. No other side effects or adverse events were reported by any of the participants.

Body weight and BMI were significantly reduced after NBS supplementation compared to the control group (p = 0.0021 and p < 0.0001, respectively). The NBS groups lost 1.568 kg after 4 weeks of treatment and the control group lost only 300 g (Table 2). When a statistical analysis of the first period was carried out, results were similar, body weight was reduced after NBS in comparison with baseline levels and when compared to the control treatment (p = 0.0004 and p = 0.0135, respectively).

No significant changes were observed in glycemia and HbA_1c levels across either treatments or between treatments. However, fasting insulin concentration was significantly reduced in NBS group in comparison with baseline values (p = 0.009), but not significant when compared to the control treatment (Table 2). A small and non-significant reduction on insulin levels was observed after control treatment (p = 0.087). As it was expected, insulin sensitivity measured by HOMA was significantly increased after NBS treatment versus baseline values. However, it was not significant when compared to the control treatment (Table 2).

The intragroup and intergroup differences for glycemia, insulin and HOMA-I variables were similar when a statistical analysis of the first period was carried out.

There were no significant effects of NBS intake on fasting triglycerides, cholesterol levels, HDL-cholesterol, LDL-cholesterol and fat body percentage. However, unexpectedly control treatment (soy milk) reduced serum triglycerides versus baseline values (p < 0.05, Mann-Whitney) and this effect was higher than the one observed in the NBS group (Mann-Whitney, p = 0.0121). No changes across either treatment or between treatments were observed on calcium, phosphate and hematological markers such as white blood cells, platelets and other indexes (not included in Table 2).