Introduction

In recent decades, the association of dietary factors with the risk of developing diabetes and increased inflammation, including high consumption of saturated fatty acids, sugary drinks, and starchy foods, combined with low consumption of fruits, vegetables, and whole grains, has been widely investigated [1]. Dietary patterns represent the totality of foods and beverages consumed by a person over a period of time; therefore, they are considered the most realistic measure of food consumption [2].

In general, nutritional studies investigating the relationship of nutrients or foods with chronic diseases, including type 2 diabetes mellitus (T2DM), encounter limitations as the human diet are complex [3]. Moreover, each food is not consumed individually, and nutrients operate in a synergistic or inhibitory manner, making it difficult to detect these possible associations [4]. This fact justifies the global consideration of the use of dietary patterns as an alternative to overcome these limitations by analyzing the effects of multiple dietary factors on individuals’ health [5].

On the contrary, evidence from cohort and prospective studies and clinical trials have shown the importance of nutrients for the prevention and control of T2DM [6, 7]. Among the various nutrients, magnesium is an essential mineral for glucose homeostasis by acting directly on the activity of glucose transporter protein 4 and the regulation of glucose entry into the cell [7]. Previous studies have observed the occurrence of magnesium deficiency in individuals with T2DM and reported its inverse relationship with metabolic control [8].

In this sense, the synergistic effects of dietary patterns must be evaluated, taking into account the nutrients, quantity, and quality of food consumed. Considering that the choice of food is of great importance for the relationship between good glycemic control and the complications associated with the disease, we hypothesized that magnesium deficiency and adherence to unhealthy dietary patterns are associated with poor glycemic control in T2DM. Thus, this study aimed to evaluate the magnesium status and dietary patterns and its relationship with glycemic control among individuals with T2DM.

Methods

Study Population

This cross-sectional study evaluated 147 individuals aged 19 to 59 years, of both sexes, with diagnostic of T2DM, and who attended primary health care in the state of Sergipe, Brazil. Participants were enrolled in the study between May 2017 and December 2019. From lists with the names and contacts of individuals with T2DM made available by health agents in primary care, the researchers contacted the individuals, and thus the exclusion criteria were screened. After this first contact, an appointment was made with the individuals on an agreed date to collect blood and other data.

Individuals who had smoking habits, consumed alcoholic beverages regularly, used vitamin-mineral supplements, were pregnant, or had other chronic diseases such as cancer, thyroid problems, arthritis, or renal diseases were excluded from the study.

This study was approved by the Ethics Committee on Human Research of the Federal University on Sergipe (opinion no. 3.012.056) and conducted according to Helsinki Declaration. All participants provided a written informed consent.

Data Collection

A questionnaire was administered to obtain the data on socio-economic characteristics, presence of diseases, use of medications, and time of T2DM diagnosis, among others. Anthropometric and body composition measurements were obtained, and the biochemical markers of glycemic and lipid control, and concentration of plasma magnesium were determined. The individuals’ food intake was evaluated to identify their dietary patterns and associate with the outcome variables.

Anthropometric and Body Composition Assessment

Weight and height were measured to obtain the BMI (kg/m2), and the results were classified according to the cutoff values established by the World Health Organization (WHO) [9]. Waist circumference (WC) was measured using an inelastic tape at the midpoint between the lower rib margin and the iliac crest, and the results were classified according to the WHO criteria [10].

The fat percentage was evaluated using a tetrapolar bioelectrical impedance scale (Biodynamics®, model 310, WA, USA). The individuals were previously instructed not to perform physical exercises and to avoid consuming coffee, tea, soda, alcoholic beverages, or chocolate a day before the test. Additionally, they were asked to empty their bladder prior to the start of the evaluation. The fat percentage was classified according to the values proposed by Lohman et al. [11].

Blood Collection

Blood samples (15 mL) were collected from the participants (after fasting for 10 h) using sterile, disposable syringes and were transferred to ethylenediaminetetraacetic acid anticoagulant tubes to determine the glycated hemoglobin percentage (%HbA1c) and to dry tubes to obtain the serum that will be used for analyzing the markers of glycemic and lipid profiles. The serum was separated from the whole blood by centrifugation at 3000 rpm for 15 min at 4 °C. Then, they were stored at – 80 °C until analysis.

Analysis of Plasma Magnesium Concentration

Initially, the samples were digested in a thermal block for 2 h at 150 °C. The procedure was performed using plasma (0.5 mL), pure nitric acid (2 mL), and nanopure water (3 mL). Then, the digested samples were diluted in nanopure water to a final volume of 15 mL, and the magnesium concentration was analyzed using the optical emission spectrometry method with inductively coupled plasma. The external calibration curves were prepared using SpecSols multielement standards at a concentration of 100 mg/L [12]. A plasma magnesium value of less than 0.75 nmol/L indicated magnesium deficiency, while that between 0.75 and 1.05 nmol/L was considered normal [13].

All reagents used were prepared with high-pressure water (resistivity: 18.2 MO cm) and obtained using a Milli-Q purification system (Millipore, USA); the material used (glassware, plastics, tips, tubes, etc.) were demineralized in a 20% nitric acid bath for at least 12 h and rinsed 10 times with nanopure water.

Analysis of Markers of Glycemic and Lipid Control

The fasting serum glucose concentration was determined using the colorimetric method, while the %HbA1c level was measured by an enzymatic method using commercial kits (Labtest, Lagoa Santa, Minas Gerais, Brazil). To evaluate the lipid profile, the serum concentrations of total cholesterol (TC), high-density lipoprotein cholesterol (HDL-c), and triacylglycerol (TAG) were determined using colorimetric methods. The low-density lipoprotein cholesterol (LDL-c) concentration was calculated using the Friedwald equation [14]. The results were classified according to the Brazilian Society of Diabetes [15] and the Brazilian Society of Cardiology [16].

Assessment of Food Intake

Food intake was assessed using three 24-h food recall (R24h) forms, which were administered on alternate days, with one day being on the weekend. The first R24h was performed in person on the day of blood collection, while the other sessions were conducted over the phone, according to the Multiple Pass Method [17].

The data obtained were analyzed using the NutWin software. This software contains data from the Brazilian Table of Food Composition, the United States Department of Agriculture table of food composition consumed in Brazil, and data from food labels provided by manufacturers. The inadequacy of the usual dietary intake of macronutrients and magnesium, calcium, zinc, and potassium were evaluated according to the acceptable macronutrient distribution ranges (AMDR) [18] and estimated average requirement and adequate intake for potassium [19], respectively.

Statistical Analyses

A descriptive analysis of the data was performed (frequency [n(%)] and mean [standard deviation—SD]). The Kolmogorov–Smirnov normality test was applied. Pearson’s or Spearman’s correlation test was performed according to the data distribution.

The usual dietary intake was assessed using the multiple source method. Then, Willet’s residual method was used to adjust the macronutrients and micronutrients by daily energy intake [20]. From the R24h data, the amounts of each food obtained through household measure were converted to grams or milliliters in the NutWin Software. The foods were listed and grouped in an Excel spreadsheet, taking into account the nutritional composition, to reduce the number of food variables, since the R24h allows an unlimited number of items to be listed.

The foods were grouped to obtain the mean absolute intake in grams/day for each food group. Thirty food groups were formed according to their similarity in nutritional composition: “fruits”; “vegetables”; “roots and tubers”; “legumes”; “beans”; “rice”; “pasta”; “poultry”; “pork”; “red meats”; “fish and seafood”; “eggs”; “sausages”; “dairy products and porridges”; “sweeteners”; “sugar”; “natural beverages”; “processed beverages”; “coffees and infusions”; “alcoholic beverages”; “whole grains and flours”; “breads, toasts, and cookies”; “breads, toast and cookies of whole grain”; “industrialized”; “sweets”; “oils and oilseeds”; “local traditional foods”; “soups and broths”; “snacks and fried foods”; and “fats.”

A priori, factor analysis was performed using a uniform sample to verify the distribution of the variables in a loading plot, contrasting the values obtained with those expected for a normal distribution. Bartlett’s test of sphericity and the Kaiser Meyer-Olkin coefficient (KMO) test were performed, and the cutoff values of P ≤ 0.05 and KMO ≥ 0.50, respectively [21, 22], indicated a satisfactory confidence level for the factor analysis.

Subsequently, principal component analysis was performed, starting with the extraction of the factors. The number of factors was selected using the Kaiser criterion (eigenvalues: > 1.0) and the scree plot analysis, through the eigenvalues plot, which explains the total variance associated with each factor [23, 24].

The factor loadings, which measure the correlations between factors, were analyzed using the varimax method of orthogonal rotation [25]. The factor loadings were either ≥ 0.25 (indicating a direct correlation with the pattern) or ≤  − 0.25 (indicating an inverse correlation with the pattern) [25]. The loading plots helped determine the factor loadings based on the location of the variables in the coordinate system. The dietary patterns were classified based on the foods that are included in each pattern.

Additionally, logistic regression models were tested to identify whether the identified dietary patterns and magnesium status were predictive of altered %HbA1c levels. Factorial scores were distributed in quartiles to assess whether higher adherence to each pattern associated with magnesium status predicted the changes in %HbA1c levels. These models were adjusted by age, sex, BMI, and time of diagnosis. A P value of less than 5% was considered significant. Model fit was verified using the Hosmer and Lemeshow test. Additionally, the multicollinearity analysis was checked using the variance inflation factor (VIF)/tolerance, being that all models presented VIFs below 2, and the tolerance measures were ≥ 0.7 (reference for VIF < 5 and tolerance > 0.2). Statistical analyses were performed using the Statistical Package for Social Science software (version 27.0).

Results

Characteristics of the Study Population

A total of 147 individuals with T2DM were evaluated, and 68% (n = 100) of them were women. The mean age and time of disease diagnosis were 48.71 (SD: 7.84) years and 6.81 (SD: 6.18) years, respectively. Most individuals had completed elementary school (59.6%) and did not practice physical activity (54.4%). The socioeconomic, lifestyle, anthropometric, clinical, and metabolic control characteristics of individuals with T2DM are shown in Table 1.

Table 1 Socioeconomic, lifestyle, clinical, body composition, and biochemical characteristics of individuals with T2DM
Full size table

The individuals evaluated had a mean BMI of 30.27 (SD: 6.14) kg/m2, and majority of them were overweight (40.1%) and obese (42.9%) (Table 1). Moreover, the individuals presented a mean WC of 99.83 (SD: 13.05) cm; approximately 9.50% and 54.40% of the women presented increased and substantially increased risk of cardiovascular diseases, respectively, while 11.60% and 12.20% of men had increased and substantially increased risk of cardiovascular diseases, respectively. Additionally, 61.2% of the participants had a risk of developing diseases associated with obesity, according to their body fat percentage.

Regarding glycemic control, 83.67% and 60.96% of the individuals evaluated presented poor glycemic control based on their fasting blood glucose values and %HbA1c levels according to Brazilian Society of Diabetes [15], respectively. Additionally, 44.90% of the individuals presented high TAG levels, 54.42% presented HDL-c levels below the reference values, and 52.38% and 70.75% presented total cholesterol and LDL-c levels above the values recommended by the Brazilian Society of Cardiology [16], respectively.

Results of the evaluation of magnesium levels in individuals with T2DM showed that 92.5% of them had magnesium deficiency, as evidenced by the mean concentration of plasma magnesium presented in Table 1.

In the evaluation of the usual dietary intake of macronutrients, the mean percentage of energy distribution for carbohydrates, lipids, and proteins was within the AMDR recommended by the Institute of Medicine [18] (Supplementary Table 1). The dietary intakes of magnesium, zinc, calcium, and potassium were deficient in most individuals assessed, as evidenced by their mean daily intake (Supplementary Table 1) and the prevalence of inadequacy in these nutrients (Supplementary Fig. 1). The prevalence rates of inadequate magnesium intake in the population studied were 87.2% in men and 82% in women. Moreover, the prevalence rates of inadequate zinc intake were 93.6% in men and 87% in women. The prevalence of inadequate calcium intake was 93.2%, while that of inadequate potassium intake was 94.6%.

When correlating micronutrient intake with glycemic control markers, potassium was inversely related to glucose (r =  − 0.165, P = 0.045) and %HbA1c levels (r =  − 0.199, P = 0.016). Magnesium intake was inversely correlated with the %HbA1c levels (r =  − 0.195, P = 0.019). The other minerals did not correlate significantly with glycemic variables (data not shown).

Food Pattern

The observed KMO value was 0.58, indicating that the sample was considered adequate for factor analysis. Through this analysis, based on Kaiser’s criteria and the scree plot, 12 components were identified (Fig. 1), and the first three factors were chosen, which were responsible for the sample variability of 28.70%.

Fig. 1
figure 1

Scree plot of the eigenvalues for each component in the factor extraction from the 24-h dietary recall

Full size image

Table 2 shows the factorial loadings of each food group obtained after varimax rotation. The first factor (named mixed) explained 13.83% of the total variance of the sample and was characterized by the consumption of beans; rice; pork; dairy products and porridge; sweeteners; natural beverages; processed beverages; coffee and infusions; bread, toast, and cookies; local habit foods; and fats. The second factor (named unhealthy) showed a variance of 7.90%. This pattern was characterized by the consumption of red meat, eggs, local habit foods, and salty and fried foods. The third factor (named healthy) showed a variance of 6.96% and was characterized by the consumption of fruits, vegetables, roots and tubers, rice, fish and seafood, eggs, coffees and infusions, and local habit foods.

Table 2 Factorial loadings obtained through Varimax rotation to obtain the dietary patterns of individuals with T2DM
Full size table

When correlating the factor loadings of dietary patterns with micronutrient intake, magnesium, calcium, and potassium were found to be inversely related to unhealthy dietary pattern (r =  − 0.477, P < 0.001; r =  − 0.488, P < 0.001; and r =  − 0.586, P < 0.001, respectively) and directly related to healthy dietary pattern (r = 0.265, P < 0.001; r = 0.372, P < 0.001; and r = 0.335, P < 0.001, respectively). Zinc was correlated only with dietary pattern mixed (r = 0.312, P < 0.001) (data not shown in tables).

As for the glycemic control variables, %HbA1c was directly correlated with the factor loadings of food patterns mixed and unhealthy (r = 0.194, P = 0.019; r = 0.181, P = 0.029, respectively) and inversely correlated with pattern healthy (r =  − 0.166, P = 0.046). Similarly, glucose concentrations were directly correlated with dietary pattern mixed (r = 0.202, P = 0.014) and inversely with pattern healthy (r =  − 0.166, P = 0.044), while pattern unhealthy showed no significant correlation (r = 0.157, P = 0.057) (data not shown in tables).

When assessing the association of dietary patterns and plasma magnesium with %HbA1c, magnesium deficiency and dietary pattern unhealthy increased the odds of elevating the %HbA1c levels by 5.893-fold (P = 0.041) and 1.550-fold (P = 0.034), respectively, after adjusting for sex, age, time of T2DM diagnosis, and BMI (Table 3).

Table 3 Logistic regression model between %HbA1c (dependent variable) and the variables plasma magnesium, dietary patterns mixed, unhealth, and health (predictor variables) of individuals with T2DM
Full size table

Corroborating the above results, we distributed the factorial scores of each dietary pattern by quartile and tested the binary logistic regression models considering the same adjustment variables as in the previous table (Table 4). In model 1, only magnesium deficiency was predictive of the changes in %HbA1c level (OR = 5.658). In model 2, magnesium deficiency was also predictive of the increase in %HbA1c level, and lower adherence to the unhealthy dietary pattern, which is characterized by the consumption of red meat, eggs, local and salty foods, and fried foods, reduced the chances of elevating the %HbA1c, as observed in Q1 and Q2, respectively (P = 0.007 and P = 0.043). By contrast, the lowest quartiles of healthy dietary pattern, which represents lower adherence to this pattern, were associated with the highest odds of increasing the %HbA1c level (Q1: P = 0.050; Q2: P = 0.044). Other models were tested to verify the association between eating patterns and glycemic and lipid variables, but the results were not significant.

Table 4 Binary logistic regression model between %HbA1c (dependent variable) and quartiles (Q) of identified dietary patterns of individuals with T2DM
Full size table

Discussion

The high prevalence of plasma magnesium deficiency in the present study sample was associated with an increased %HbA1c level. In parallel, three main eating patterns were identified and represent the dietary habits of individuals with T2DM, of which unhealthy eating pattern, characterized by the consumption of red meat, eggs, local foods, and salty and fried foods along with magnesium deficiency, was associated with the highest chance of increasing the %HbA1c level. Furthermore, lower adherence to the healthy dietary pattern, characterized by consumption of vegetables, roots, tubers, rice, fish and seafood, eggs, and local foods, was also associated with poorer glycemic control.

Serum magnesium concentrations are inversely associated with glycemic markers (glucose, fasting insulin, and %HbA1c) [26] and BMI [27]. Other studies have observed negative associations between serum magnesium concentration and insulin resistance, and positively with insulin sensitivity [26, 28]. On the contrary, poor glycemic control reduces the serum concentrations of this nutrient by affecting the tubular reabsorption of magnesium [26, 29]. Studies relating both dietary patterns and micronutrient-related nutritional status with glycemic markers in the T2DM population are limited, especially with nutrients that are involved in metabolic control. In addition, these micronutrients are also components of the foods that comprise dietary patterns known to be beneficial for reducing the risk of metabolic diseases, such as the DASH dietary pattern, which promotes the reduction of blood pressure by encouraging the consumption of foods low in saturated fat, total fat, cholesterol, and sodium and high levels of potassium, calcium, magnesium, fiber, and protein [30].

Dietary patterns have been studied to promote a greater understanding of the relationship between diet and diseases, especially chronic noncommunicable diseases [31], as their effects are related to the interaction between the foods that comprised each pattern. Thus, different dietary patterns can positively or negatively affect the metabolic control of individuals with T2DM [32].

Similar to the findings of the present study, Dekker et al. [33] found some associations between the “meats and snacks” eating patterns with increased %HbA1c and fasting blood glucose levels. Another study showed that %HbA1c level was also elevated among individuals with T2DM who had a higher intake of cereals, vegetables, and meats compared with those with moderate intake of cereals and meats, high intake of vegetables, low intake of cereals, and moderate intake of vegetables and meats [34].

The increase in %HbA1c level associated with dietary patterns that involved the consumption of meat can be explained by the high amounts of saturated fat that is directly associated with hyperinsulinemia and the increased level of insulin resistance [35, 36]. Additionally, when this type of food is cooked, the formation of advanced glycation end products that can increase the production of pro-inflammatory cytokines, such as tumor necrosis factor-alpha (TNF-α) and interleukins, which are related to increased complications of T2DM [37, 38].

Unhealthy dietary patterns are also associated with increased levels of circulating lipids, as observed by Buscemi et al. [39] who reported that an unhealthy eating pattern (high consumption of soft drinks, fried foods, seed oils, cured meats, butter, red meats, and sweets) was associated with a higher TAG/HDL-c ratio than the other patterns evaluated (the Mediterranean dietary pattern and intermediate dietary pattern). In the present study, no associations were observed between the identified food patterns and the variables of lipid control in individuals with T2DM.

The consumption of foods that make up the dietary pattern recommended by “MyPlate” for individuals with T2DM in the Arab Emirates was associated with lower %HbA1c level; the consumption of fresh fruits promoted glycemic control, while the consumption of soft drinks and fast foods at least once a week increased the risk of poor metabolic control. That study also found that fruit juice intake more than once a day was predictive of poor glycemic control [40].

The lower adherence to the healthy dietary pattern was also associated with higher odds of increasing the %HbA1c level to > 7%. In a study by Sarmento et al. [41], the healthy eating patterns (characterized by consumption of fruits, vegetables, whole-wheat cereals, dairy products, low-fat meats, and fish) were associated with a reduction in fasting plasma glucose, %HbA1c, and LDL-c cholesterol levels. In another study conducted in Brazil, a similar pattern called “healthy pattern” (characterized by consumption of natural fruits and juices, olive oil and oilseeds, roots, tubers, and vegetables) showed no correlation with serum glucose concentrations [22]. However, they observed a negative correlation between traditional Brazilian food patterns (characterized by consumption of rice, beans, and chicken) and serum glucose concentrations.

The composition of foods that characterize the healthy dietary pattern may have favorable effects on glycemic control due to the marked presence of fiber, antioxidant substances, unsaturated fatty acids, and micronutrients such as magnesium, zinc, and potassium [42,43,44]. These nutrients improve glycemic control by reducing postprandial glycemia [42] and %HbA1c levels [45] as well as by improving insulin resistance [42]. Moreover, these foods act positively on satiety [46]. The T2DM individuals assessed in this study showed a high prevalence of inadequate intake of magnesium, zinc, calcium, and potassium; the %HbA1c levels correlated negatively with the dietary intake of magnesium and potassium, while the serum glucose concentration correlated negatively with the intake of potassium. Previous studies have pointed out that these minerals are associated with better glycemic control; however, they have been studied separately, and individuals with T2DM has less intake of these minerals in most of these studies [47,48,49]. Brandão-Lima et al. [50] observed that reduced intake of zinc, calcium, magnesium, and phosphorus was associated with a greater chance of increasing the %HbA1c level to > 7% in individuals with T2DM than in those with higher intake of this mineral.

However, the use of R24h to assess the nutrient intake has certain limitations such as the need for the individual’s memory to accurately report the foods and portions consumed, and the frequency of underreporting by overweight individuals. However, to reduce the bias, the intra- and inter-individual variabilities and adjustment of micronutrient intake by energy intake were performed. Notably, the strength of this study was the inclusion of results of the plasma magnesium concentration concomitant with the analysis of dietary patterns since this nutrient plays a strong role in insulin action and in promoting glycemic control and few studies have been conducted with this dual approach. It is noteworthy that this approach including micronutrient assessment combined with dietary patterns is not often observed in studies. This suggests new perspectives for evaluating specific dietary patterns that promote better glycemic control, as seen in the Dash dietary pattern that includes micronutrients that favor blood pressure control.

Conclusion

We conclude that the plasma magnesium deficiency and adhesion to a dietary pattern consisting of red meat, eggs, regional foods, salty, and fried foods, considered unhealthy, were associated with poor glycemic control. Moreover, this dietary pattern also was directly related to a lower intake of magnesium, calcium, and potassium due to the composition of the foods that characterize it. Lower adherence to the healthy dietary pattern and the presence of concomitant plasma magnesium deficiency was also associated with poor glycemic control, since the foods that characterize this pattern (fruits; vegetables; roots and tubers; rice; fish and seafood; eggs; coffees and infusions; and local habit foods) present good nutrient density, for rich in micronutrients such as magnesium, and are low in calories, and relate to better diabetes control. These results show that for better glycemic control, it is necessary to adopt eating patterns that contain foods that can interact in a beneficial way, improving the nutritional status of micronutrients, given the high prevalence of inadequate intake of magnesium, zinc, calcium, and potassium, as well as plasma magnesium deficiency.

Therefore, a combined approach that explains such relationships strengthens the understanding of the interactions between the nutrients that make up dietary patterns and the synergy between them. Thus, studies of this nature that aim to understand the interactions between dietary patterns and nutritional status related to micronutrients are important for the development of nutritional strategies aimed at the metabolic control of individuals with T2DM, favoring the reduction of the risk of complications associated with the disease and improvement in the quality of life of this population.