The Effect of the Macrobiotic Ma-Pi 2 Diet vs. the Recommended Diet in the Management of Type 2 Diabetes

The Randomized Controlled MADIAB Trial

Andreea Soare; Yeganeh M Khazrai; Rossella Del Toro; Elena Roncella; Lucia Fontana; Sara Fallucca; Silvia Angeletti; Valeria Formisano; Francesca Capata; Vladimir Ruiz; Carmen Porrata; Edlira Skrami; Rosaria Gesuita; Silvia Manfrini; Francesco Fallucca; Mario Pianesi; Paolo Pozzilli


Nutr Metab. 2014;11(39) 

In This Article


Trial Design

The study was a 21-day, controlled open-label trial in which participants were randomized (1:1 ratio) to the Ma-Pi 2 macrobiotic diet or to a diet based on dietary recommendations guidelines for type 2 diabetes.[25] The trial took place in the spring of 2013 at 2 closed site hotels in Italy. Throughout the trial, participants stayed at two different hotels according to the type of diet they were randomized to. The hotels were localized in the same geographic area, very close to each other, approximately a 20 minutes' drive distance between them. Patients were recruited from the Endocrinology and Diabetes Department at University Campus Bio-Medico. The trial was conducted in accordance with the Declaration of Helsinki and the Good Clinical Practice guidelines and approved by the Institutional Review Board of University Campus Bio-Medico.

Inclusion and Exclusion Criteria

Overweight or obese (BMI 27–45 kg/m2) males and females, aged 40–75 years affected by type 2 diabetes were recruited by the medical team during regular visits to the Department of Endocrinology and Diabetes of the University Campus Bio-Medico in Rome, Italy. Additional inclusion criteria were a diagnosis of type 2 diabetes at least 1 year prior to the start of the trial and management with dietary intervention, oral hypoglycemic drugs (OADs), or both for 6 months prior to study entry. Identification of the presence of associated metabolic syndrome was made in accordance to the National Cholesterol Education Program Adult Treatment Panel III criteria, even though it was not an inclusion criteria.[26] Exclusion criteria included current use of insulin, or use at any time in the 2 years prior to the study, current use of corticosteroid therapy or any other drug that could interfere with carbohydrate metabolism, alcohol abuse, pregnancy, or already following a macrobiotic diet.


Participants' eating habits prior to study entry were assessed using qualitative and quantitative questionnaires, and the energy and nutritional content of their diet was evaluated based on the answers given. Group assignment to treatment was blinded for personnel involved in collection of follow-up data and for personnel analyzing blood samples until data was locked for statistical analysis.

Participants were randomized to either the Ma-Pi 2 diet or a control diet according to the dietary guidelines for type 2 diabetes. The Ma-Pi 2 diet consisted of whole grains, vegetables and legumes. Beicha tea (roasted green tea) represented the main source of liquids, while the control diet was adapted to the Mediterranean culinary style. For both groups, energy intake was restricted by limiting calories to 1900 kcal/day and 1700 kcal/day for males and females, respectively. The diets were isocaloric but differed in nutrient composition. Ma-Pi 2 diet derived 72% of energy from carbohydrate, 18% from fat, and 10% energy from protein, fiber equal to 30 g/1000 kcal, while the control diet 50% from carbohydrate, 20% from protein, and 30% from fat, fiber ≥20 g/1000 kcal. Alcohol consumption was forbidden. Both diets provided 5 meals per day, with energy intake being divided between meals, 20% calories at breakfast, 30% calories at lunch and 30% calories at dinner. Two snacks were administered at approximately 2.5 hours after breakfast and lunch, respectively, each contributing 10% of the calories per day (Additional file 1

A 10-days menu cycle was devised for both diets and then repeated at the end of the period for another 11 days. Each daily menu was carefully planned and completed with recipes for each dish. Nutritional analysis and menu planning was developed with MètaDieta® Software using the Italian Food Composition Tables edited by the National Institute for Food and Nutrition Research (INRAN).[27] Each menu cycle was ready a week before the beginning of the study to ensure that each recipe was tested by the cooks in order to reproduce it consistently. At this stage and throughout the intervention study, dietitians checked and weighed each portion size before and after cooking to make sure that subjects in both groups consumed equivalent amounts of energy and that macronutrient content was respected. Each subject was informed that a plate waste greater than or equal to 5% of the total daily food amount meant dismissal from the trial. Throughout the trial, participants stayed at two different hotels according to the type of the diet they were randomized. Meal consumption was strictly controlled in order to evaluate dietary compliance during the trial. Participants had their meals together with physicians and dietitians who could check their adherence to study protocol. Dietary adherence was defined as absence of any transgression from the assigned diet. Those subjects who attended meal session for fewer than 20 out of 21 days were considered non-adherent.

Participants were asked not to alter their exercise habits during the intervention period. Physical activity was assessed daily using a pedometer (Tri-axial activity monitor XL-18/XL-18 CN-AND A&D Medical-California-USA). In addition they were instructed to continue their pre-study OAD doses without modification throughout the study, unless hypoglycemic symptoms were accompanied by a capillary glucose reading <70 mg/dL. In such cases hypoglycemic medications were reduced for participant safety.

Outcome Measures

The primary outcomes of this study were the percentage change in Fasting Blood Glucose (FBG) and Post Prandial Blood Glucose (PPBG) levels from baseline (T0) to the 21st day of treatment (T21) in Ma-Pi 2 group compared to the control group. Secondary outcomes included percentage change from baseline in plasma concentrations of HbA1c, total cholesterol, LDL cholesterol (LDLc), HDL cholesterol (HDLc), LDL/HDL ratio, and percentage change from baseline of insulin resistance, body weight, BMI, waist and hip circumference, and number of patients who achieved target values of FBG ≤110 mg/dl and PPBG ≤140 mg/dl.

FBG and PPBG were measured daily from T0 to T21 by the medical staff using capillary blood glucose meters (Bluecare, Biochemical Systems International, Arezzo, Italy). The fasting blood glucose was measured right before meals and the post-prandial blood glucose was measured 2-h after lunch. For all participants, venous blood samples were obtained early in the morning after a 12 hour fasting period. All biochemical and anthropometrical measures were assessed at T0 and T21 by the central laboratory (University Campus Bio-Medico, Rome). Insulin resistance was calculated using the homeostasis model assessment of insulin resistance (HOMA-IR).[28]

Body weight was measured at T0 and T21 before breakfast using a digital scale accurate to 0.1 kg, waist circumference was measured with a tape measure placed 2.5 cm above the umbilicus. Hip circumference was measured at the maximal protrusion of the buttocks.

All patient and investigator-reported adverse events (AEs) were recorded at each visit.

Statistical Analyses

This study required 28 randomized patients per group with 80% power to detect a difference of at least 12 percentage points reduction in mean FBG and PPBG from baseline between Ma-Pi 2 and control groups (with a 2-sided type I error at 0.05), assuming a standard deviation of ≤15% and a maximum dropout rate of 11%.

To detect a 0.25 percentage point between-group difference in HbA1c from baseline (the main secondary endpoint) with 80% power, a p value at the 5% level with an assumed SD of 0.3% and 10% estimated withdrawal, 27 participants were required per group.

The primary analysis was based on a modified intention-to-treat principle and it was carried out for all patients who had FBG and PPBG results for at least the first week following randomization. A non-parametric approach was chosen for the statistical analysis since outcome variables were found to be of non-normal distribution (using the Shapiro test). Quantitative variables were summarized using percentiles (median and interquartile range). Comparisons between treatment groups were performed using the Wilcoxon rank-sum test and 95% confidence intervals (95% CI) for median values. Absolute and percentage frequencies were used for qualitative variables and the Fisher exact test was applied for group comparisons.

The 2 groups were compared at T0 to test whether patients were similar in terms of demographic and anthropometrics characteristics, and lipid and carbohydrate metabolic parameters. The percent variations between values measured at T0 and T21 were calculated as efficacy variables for the primary (FBG and PPBG), and secondary end-points (HbA1c, HOMA-IR, total cholesterol, LDLc, HDLc, LDL/HDL ratio, triglycerides, BMI, waist and hip circumference).

A bivariate analysis was performed to compare measures changes between the Ma-Pi 2 diet group and the control group and results were graphically represented by means of box-plots (Figure 1). A linear quantile regression analysis[29] was performed to estimate the effect of the diet (Ma-Pi 2 vs. control, explanatory variable) on the median percentage change occurred in each measured variable as (dependent variables), between T0 and T21. Each model was adjusted for those variables that could potentially affected the percent changes in the dependent variables, i.e. gender, age, BMI at baseline, and physical activity done by subjects during 21 days of the treatment (measured as the median number of kilometers walked per day). When HOMA-IR was analyzed, the wrist circumference was included in the model as covariate for adjustment. Wrist circumference is recognized a valid and easy marker to measure insulin resistance.[30,31] The results of linear quantile regression analysis were expressed as point and interval estimates of regression coefficients; when the coefficient was positive, the measured variable reduction was related to Ma-Pi 2 diet effect and when the coefficient was negative the reduction was related to the control diet. When the 95% CI's did not include zero, regression coefficients were considered statistically significant (Table 1).

Figure 1.

Reduction in primary and secondary outcomes from baseline to T21 in the Ma-Pi 2 group vs. controls. Primary outcomes - Percentage change in FBG and PPBG from baseline to T21 in the Ma-Pi 2 group vs. control group. The figure represents the results of the bivariate analysis when the Ma-Pi 2 and control groups were compared in the percentage change between T0 and T21 of the efficacy variables (non-adjusted values). Secondary outcomes - Percentage change in HbA1c, lipid panel, BMI, weight, hip circumference and waist circumference from baseline to T21 in the Ma-Pi 2 group vs. control group. The figure represents the results of the bivariate analysis when the Ma-Pi 2 and control groups were compared in the percentage change between T0 and T21 of the efficacy variables (non-adjusted values).

Patients with values of FBG ≤110 mg/dl and of PPBG ≤140 mg/dl were considered to have reached target values. Group comparisons in terms of percentage of patients achieving and maintaining those target levels after 21 days were performed using McNemar's test to estimate the differences between proportions in the 2 groups and their relative 95% CI's; when the 95% CI's did not include zero, the differences were statistically significant (Table 2).

A ranked-based nonparametric analysis for longitudinal data was used to compare the time trend in absolute values of FBG and PPBG over 21 days of dietary intervention in Ma-Pi 2 and control groups. The effect of diet, time and their interaction were evaluated; results were graphically represented by plotting the median values of the daily measurements of FBG and PPBG in the two groups (Figure 2).

Figure 2.

Linear distribution of the daily blood glucose values during the 21 days in and between groups. The figure represents the results of a non-parametric longitudinal analysis of all daily FBG and PPBG values between baseline and T21 in the Ma-Pi 2 group and the control group.

All statistical analyses were performed using R statistical package (Foundation for Statistical Computing, Vienna, Austria) and statistical significance was assessed at a level of probability of 0.05.