Prediabetes is characterised by elevated blood glucose levels above the normal range, but below the diagnostic threshold for type 2 diabetes mellitus (T2DM)(1). While T2DM is characterised by insulin resistance and partial deficiency of insulin secretion by pancreatic β cells(2). Both prediabetes and T2DM are associated with obesity and ageing and represent an increased risk of all-cause mortality and CVD in the general population and in patients with CVD(Reference Cai, Zhang and Li3). In 2019, 463 million adults between the ages of 20 and 79 years had diabetes, 90 % of whom were T2DM. Globally, it is estimated that 578 million people will have T2DM by 2030(Reference Saeedi, Petersohn and Salpea4). In the last three decades, Brazil has become the fourth country with the highest number of people with T2DM, contributing to the rapid growth of the disease burden in the country(Reference de Almeida-Pititto, Dias and de Moraes5–Reference Duncan, Schmidt and Ewerton7).
T2DM in adolescence is becoming an important global public health problem in the 21st century, largely affecting high-income countries, but also medium and lower income due to the epidemiological and nutrition transitions(Reference Viner, White and Christie8). T2DM in adolescence often progresses rapidly, presenting a risk of complications that impact quality of life at relatively young ages(Reference Lascar, Brown and Pattison9,Reference Grøndahl, Johannesen and Kristensen10) . The Study of Cardiovascular Risks in Adolescents (ERICA), carried out in Brazil, reported a prevalence of prediabetes of 22 % and T2DM of 3·3 % in the age group of 12–17 years(Reference Telo, Cureau and Szklo11).
Physical activity and diet can prevent or delay the onset of T2DM(Reference Knowler, Barrett-Connor and Fowler12). In terms of dietary composition, dairy products are nutritionally dense foods with scientific support for cardiometabolic benefits(Reference Rojo-Martínez13). Several studies point to an inverse association between dairy products consumption and the risk of prediabetes and T2DM in adults(Reference Gijsbers, Ding and Malik14–Reference Slurink, Voortman and Ochoa-Rosales16). Reduction in risk of T2DM from dairy products intake may be due to dairy products components, including Ca, Mg, K, trans-palmitoleic fatty acids and low glycaemic load(Reference Pittas, Dawson-Hughes and Li17–Reference Comerford and Pasin19). Evidence also indicates that dairy products may promote beneficial changes in the gut microbiota and have positive association with subclinical inflammation(Reference Hirahatake, Slavin and Maki20), pathogenic mechanisms involved in the genesis of T2DM(Reference Bock, Telo and Ramalho21,Reference Riboldi, Luft and Bracco22) .
Although the inverse association between dairy products consumption and T2DM risk has been widely documented(Reference Soedamah-Muthu and de Goede23–Reference Gao, Ning and Wang25), the role of specific types of dairy products (e.g. fat composition and fermentation) remains unclear. Yogurt consumption might have a protective effect on inflammation, while certain types of cheese may be associated with a pro-inflammatory status(Reference Gadotti, Norde and Rogero26). However, the association between dairy products consumption and glycaemic metabolism and T2DM has been poorly studied among adolescents. Considering that dairy products are an important food group in the diet of adolescents, it is essential to understand their impact on chronic non-communicable diseases. Nevertheless, there is a lack of research on this subject concerning adolescents, and no representative studies have been conducted in Brazil until this moment.
The present study aimed to evaluate the association between dairy products consumption and its subtypes and likelihood of T2DM in Brazilian adolescents and to describe the association between total dairy products and its subtypes and measures of fasting glucose, glycated hemoglobin (HbA1c), and insulin resistance (HOMA-IR) in adolescents.
Methods
Study design
ERICA is a multicentre, cross-sectional, school-based, nationwide study that aimed to estimate the prevalence of cardiovascular risk factors in a representative sample of Brazilian adolescents aged 12–17 years who attended public and private schools in Brazilian municipalities with more than 100 000 inhabitants. The study design was based on complex sampling to obtain representativeness at national, regional and capital levels. The sampling involved thirty-two strata composed of the twenty-seven capitals and another five groups of cities with at least 100 000 inhabitants from each Brazilian macro-region, totalling 124 municipalities in the country. In these cities, 1247 public and private schools were selected. In each school, three classes were selected per school with different combinations of school hours (morning and afternoon) and grade (seventh, eighth and ninth grades of elementary school and first, second and third grades of high school). All students from selected classes were invited to participate.
This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects/patients were approved by the Research Ethics Committee of the Federal University of Rio de Janeiro and by the Ethics Committees of each of the twenty-seven states and the Federal District. Written informed consent was obtained from all subjects, and their parents or guardians signed a consent form to participate in the study. More information on ERICA’s design and sampling can be obtained from previous publications(Reference Bloch, Szklo and Kuschnir27,Reference Vasconcellos, Silva and Szklo28) .
Of the 102 327 adolescents eligible to participate in ERICA, 72 508 were morning students and, therefore, able to participate in the blood collection. The study included 36 956 adolescents who had participated in blood sampling, provided data through a questionnaire, and completed a 24-h dietary recall. The flow chart of eligible students and the total sample analysed in the present study are shown in online Supplementary Fig. 1.
Dietary assessment
Dietary intake was assessed with a 24-h recall through a face-to-face interview conducted by trained evaluators. In addition, a random subsample of two students per class (10 % of the total sample) was selected to respond to a second 24-h recall. The recalls were collected using the multiple-pass method. All food and beverages consumed the day before were recorded using photographs to estimate portion sizes. The ingredients and reported foods were compared from a standardised list of foods or added to the software developed specifically for ERICA (ERICA-REC)(Reference Barufaldi, Abreu and Veiga29), and the conversion of foods into grams was performed according to the table of standard measures for foods consumed in Brazil(30). Subsequently, for information on the consumption of macro and micronutrients, the data were linked to the table of the nutritional composition of foods consumed in Brazil(31). Finally, the Multiple Source Method programmes and the National Cancer Institute (NCI) method for foods consumed episodically were used to correct for intra-individual variability and allow, with some degree of uncertainty, to calculate the correction factor for dairy products, macro and micronutrients between the first and second R24h and thus estimate the usual food intake of each adolescent(Reference Freedman, Guenther and Dodd32,Reference Kipnis, Midthune and Buckman33) .
To assess dairy products consumption in the present study, two independent researchers evaluated all food items reported in the 24-h recall and classified dairy products and preparations (e.g. desserts) with added milk. Milk-based preparations were considered those whose milk content was ≥ 50 ml per recipe. Subsequently, dairy products were classified into the following groups: total dairy products (all types of dairy products were included), full-fat dairy products, those containing about 3·5 % fat (whole milk, yogurt and cheese, milk drinks, chocolate drinks such as Toddynho ® and Chocomilk ® , butter, curds, cream of milk, desserts containing ≥ 50 ml of milk, including milk caramel, pudding and ice cream), low-fat dairy products, those containing about ≤ 2 % fat (milk, cheese, low-fat yogurt, and diet or light milk drinks), milk (all types), yogurt, cheeses, butter, desserts and fermented foods (all types of yogurt and cheese, fermented milk and Yakult ® ). The food items sour cream and curd were not analysed separately due to their low consumption among the population studied.
Daily dairy products consumption was calculated in grams and then categorised into servings per d (sv/d). The gram equivalent of the servings was 240 g for milk, 120 g for yogurt, 30 g for cheese, 5 g for butter, 50 g for milk-based desserts and 75 g for fermented dairy products(Reference Drehmer, Pereira and Schmidt34).
Outcome: prediabetes and diabetes
Participants were instructed to fast for 12 h before the exam. Fasting blood samples were collected for analysis of glucose, HbA1c and homeostatic model assessment-insulin resistance (HOMA-IR) (glucose (mmol/l) × insulin (mU/l)/22·5]. Blood samples were processed within 2 h of collection and were kept between 4°C and 10°C while being transferred to the single-study laboratory(Reference Bloch, Szklo and Kuschnir27). Blood collection for exams was conducted early in the morning after an overnight fasting, and the 24-h dietary recall was administered on the day after the blood sampling or later. This approach tried to ensure that fasting did not interfere on the diet report(Reference Bloch, Szklo and Kuschnir27). For the present study, those who had a previous diagnosis of diabetes or who reported using insulin/oral antidiabetic drugs were excluded, because these adolescents could already have had previous changes in their diet, which would influence the studied outcomes (n 1342). To estimate prevalence of prediabetes and undiagnosed T2DM, fasting glucose and HbA1c measurements were used, following the diagnostic criteria of the American Diabetes Association(35). Prediabetes was defined by fasting glucose levels between 100 and 125 mg/dl (5·6–6·9 mmol/l) or HbA1c between 5·7 % and 6·4 %. T2DM was defined by fasting glucose ≥ 126 mg/dl (7·0 mmol/l) or HbA1c ≥ 6·5 %.
Covariates
The covariates investigated were sex, age (categorised as 12–13, 14–15 and 16–17 years), self-reported skin colour (brown, white, black, yellow and indigenous), type of school (public or private), macro-region (North, Northeast, Midwest, Southeast and South) and frequency of breakfast consumption (never, sometimes or always) and smoking (yes or no). Other covariates are BMI, waist circumference, physical activity, screen time, energy intake and selected food groups and nutrients, which will have their measurement methods described below.
Adolescents’ height was measured using a calibrated portable stadiometer (Alturexata®, Minas Gerais) with the students barefoot and with their heads in the Frankfurt plane position. Weight was measured using a digital scale (model P150m, 200 kg capacity, and 50 g precision, Líder®)(Reference Bloch, Szklo and Kuschnir27). BMI was calculated [BMI = weight (kg)/height (m)2] and classified according to the WHO reference(Reference de Onis, Onyango and Borghi36) using sex-and age-specific cutoffs. The categories were as follows: malnutrition Z-score < −3; low weight ≥ −3 and < −1; normal weight ≥ −1 and ≤ 1; overweight > 1 and ≤ 2; and obesity > 2. Waist circumference was measured at the midpoint between the iliac crest and the last rib with the students standing with the abdomen relaxed at the end of a gentle exhalation. All measurements were made in duplicate for quality control purposes(Reference Bloch, Szklo and Kuschnir27).
Moderate to vigorous physical activity was assessed using an adapted version of the Self-Administered Physical Activity Checklist, cross-culturally adapted and validated in Brazilian adolescents(Reference Farias Júnior, Lopes and Mota37). The duration and frequency of physical activities practiced in the last 7 d were multiplied to determine the weekly time of physical activity and then dichotomised into < 60 (inactive) or ≥ 60 (active) min/d(Reference Bull, Al-Ansari and Biddle38). For smoking, the ERICA questionnaire asked whether adolescents smoke (yes or no) and, for screen time, the question was how many hours adolescents use the computer, watch TV, or play video games, and this variable was dichotomised (≤ 2 h per d or > 2 h per d).
In addition to dairy products consumption, other food groups were considered as covariates in the analyses: whole grains, vegetables, fruits, processed meat, snacks and sweets and sugary drinks, fruit juices, and alcoholic beverages(Reference Ronca, Blume and Cureau39). Energy and nutrient intake (SFA and PUFA and Ca) were also evaluated.
Statistical analysis
The analyses were performed using Stata software (version 14), considering the complex sample design. Usual consumption estimates were calculated using SAS software (version 9.4).
The variables related to food consumption (24-h recall) were estimated using the National Cancer Institute method, which corrects for intrapersonal variability (daily variation), via SAS macros. A statistical model was applied that allows estimating the usual distributions of food intake consumed episodically based on a two-part non-linear mixed model: the first part uses logistic regression with a person-specific random effect to estimate the probability of consuming a meal; the second part uses linear regression on a transformed scale with a specific person effect to specify the amount of consumption/d.
Multivariable linear regression models were used to test the association between daily servings of dairy products (total and subtypes) and the following metabolic variables: fasting glucose, HbA1c and HOMA-IR. Linear trend was tested by modelling dairy products servings per d as continuous variables. Poisson regression with robust multivariate variance was used to assess the association between dairy products consumption (total and subtypes) and the combined prevalence of prediabetes and T2DM. We did not conduct analyses for prediabetes and T2DM separately due to the limited number of adolescents with T2DM and who reported consuming specific types of dairy products. This would have led to imprecise estimates, making it difficult to draw meaningful conclusions.
All analyses were performed for the total sample and stratified by BMI. Model 1 was adjusted for sex, age, skin colour, school type and macro-region. Model 2 included model 1 and the variables skipping breakfast, physical activity, total energy consumption, food groups (fruits, vegetables, whole grains, beans and legumes, processed meat, snacks, sweets, and sugary drinks), smoking, and screen hours. Finally, model 3 included all variables from model 2 plus BMI in kg/m2, to investigate their specific role in the association.
Results
In total, the final sample consisted of 35 614 adolescents. In online Supplementary Table 1, the description of the final sample is presented. Females were 60·1 % of the sample, 36·4 % were between 14 and 15 years old, 52·9 % reported brown skin colour, and 73·8 % were students from public schools. The prevalence of obesity was 8 %, and overweight was 17·5 %. Regarding physical activity, 53·3 % were in the inactive category. We identified 4870 (13·7 %) new cases of prediabetes and 83 (0·23 %) new cases of T2DM.
Table 1 shows the characteristics of the sample related to sociodemographic, nutritional, behavioural and anthropometric variables according to the number of servings of dairy products consumption per d. Dairy products consumption was higher in males, in adolescents between 16 and 17 years of age, in those who reported brown skin colour and among adolescents from public schools. According to region, higher consumption of dairy products was observed among adolescents in the Northeast, among those who always ate breakfast, among those with a normal BMI, who were within the active category, among those who consumed more energy content a day and reported more from 2 h per d in screen time. Regarding the consumption of food groups, the highest consumption of dairy products was observed among adolescents who consumed more fruits, whole grains, beans and vegetables, processed meat, snacks and sweets, and sugary drinks.
Table 1. Demographic, socio-economic, nutritional, behaviour and metabolic characteristics according to dairy products consumption categories of Brazilian adolescents without previously diagnosed diabetes. ERICA* (n 35 614)
* ERICA: Study of Cardiovascular Risk in Adolescents.
† BMI
‡ Normal weight includes normal (Z-scores ≥ –1 e ≤ 1), low weight (Z-scores ≥ –3 and < –1) and malnutrition Z-scores < –3 (OMS 2017).
§ Whole grains: cereals with no added sugar, whole grain bread, granola, oats, flaxseed, amaranth; baked potato, wholemeal pasta, brown rice, yam, cassava, yam and maize.
|| Processed meat: fried meats, nuggets, sausage, ham and salami.
¶ Snacks and candies: all types of sweet and savoury cookies, cereal bars, powdered chocolate (Nescau ® and similar), jellies, chocolates, oilseeds with sugar or chocolate (dessert).
** Sugared drinks, fruit juices and alcohol: natural fruit juices, box juices, powdered juices (Tang ® and Clight ®), energy drinks, vodka and beer.
†† SFA.
‡‡ PUFA.
§§ HbA1c: glycated hemoglobin.
|||| HOMA-IR: homeostatic model assessment-insulin resistance.
Total dairy products intake and dairy subgroups, in servings per d, are shown in Table 2. There was no difference in consumption between the overall ERICA sample and the final sample analysed. Intake of full-fat dairy was higher than intake of low-fat dairy. In our analysis of daily dairy products consumption, we found that milk and butter contributed the most grams per d, followed by cheese. Online Supplementary Table 2 shows the median of dairy products consumption groups.
Table 2. Total intakes of dairy products and dairy products subgroups for the final sample of participants: ERICA* (n 35 614)
* ERICA: Study of Cardiovascular Risk in Adolescents.
† Whole milk, low-fat milk, yogurt (regular and low-fat), cheese (regular and low fat), butter, desserts made with milk and fermented dairy products (cheese, yogurt, fermented milk and curd).
Table 3 shows a significant inverse association between the highest total consumption of dairy products and fasting blood glucose levels (models 2 and 3) in the total sample, after adjustment for behavioural, nutritional and anthropometric variables. Likewise, there was an inverse association between dairy products consumption and HOMA-IR in the total sample only after adjusting for demographic, behavioural and nutritional data (models 1 and 2), with attenuation by BMI (model 3). When stratified according to BMI, we observed that the inverse association remained significant for fasting glucose and HOMA-IR only in overweight or obese adolescents. We did not find a linear association between total dairy products consumption and HbA1c.
Table 3. Multiple linear regression for fasting glucose, HbA1c and HOMA-IR by categories of total dairy products intake for the total sample and by BMI category, ERICA
FG, fasting glucose; HbA1c, glycated hemoglobin; HOMA-IR homeostatic model assessment-insulin resistance.
Values are adjusted by multivariable linear regression. Values are β (95 % CI).
Linear trend was tested by modelling dairy products servings per d as a continuous variable in the multivariable regression models (P-value).
P-values for interaction between total dairy products intake and BMI for each outcome: FG = 0·006; HbA1c = 0·584; HOMA-IR < 0·001.
* FG, HbA1c and HOMA-IR as continuous outcomes.
† Normal weight includes normal (Z-scores ≥ –1 e ≤ 1), low weight (Z-scores ≥ –3 and < –1) and malnutrition Z-scores < –3 (OMS 2017).
‡ Model 1: adjusted for age, sex, region (North, Northeast, Midwest, Southeast and South), skin colour and type of school.
§ Model 2: adjusted for model 1 plus breakfast, physical activity, energetic intake, food groups (fruits, vegetables, whole grains, beans and legumes, processed meat, snacks and sweets, sugary drinks), smoking and screen time.
|| Model 3: adjusted for model 2 plus BMI.
Regarding the dairy products subgroups (whole fat and low fat), an inverse association was found between higher full-fat dairy products consumption and HOMA-IR in models 1 and 2. Upon BMI stratification, an inverse relationship was observed between higher consumption of full-fat dairy products and fasting blood glucose and HOMA-IR, after adjusting for confounders in the overweight or obese sample (Table 4).
Table 4. Multiple linear regression for fasting glucose, HbA1c and HOMA-IR‡ by dairy products subgroups (whole and low-fat dairy products) for the total sample and by category of BMI
FG, fasting glucose; HbA1c, glycated hemoglobin; HOMA-IR homeostatic model assessment-insulin resistance.
Values are adjusted by multivariable linear regression. Values are β (95 % CI).
Linear trend was tested by modelling dairy products servings per d as a continuous variable in the multivariable regression models (P-value).
P-values for interaction between dairy products subgroup and BMI for each outcome: whole dairy products, FG = 0·002; HbA1c = 0·201; HOMA-IR = 0·009; low-fat dairy, FG = 0·321; HbA1c = 0·726; HOMA-IR = 0·738.
* ERICA sample low consumption: ≤ Median (P50).
† ERICA sample high consumption: > Median (P50).
‡ FG, HbA1c and HOMA-IR as continuous outcomes.
§ Model 1: adjusted for age, sex, region (North, Northeast, Midwest, Southeast and South), skin colour and type of school.
|| Model 2: adjusted for model 1 plus breakfast, physical activity, energetic intake, food groups (fruits, vegetables, whole grains, beans and legumes, processed meat, snacks and sweets, sugary drinks), smoking and screen time.
¶ Model 3: adjusted for model 2 plus BMI.
There were no clear associations between total dairy products consumption and the prevalence of combined prediabetes and diabetes in the total sample. But, when we stratified according to BMI, we observed that the consumption of 2–3 servings per d of dairy products reduced 19·5 % of prediabetes and T2DM prevalence ratio (PR 0·80 95 % CI 0·65, 0·98), only in overweight or obese adolescents. (Table 5).
Table 5. Poisson regression of the association between prediabetes and diabetes detected in the study and consumption of total dairy products category for the total sample and by BMI category
PR, prevalence ratio.
Values are adjusted PR and 95 % CI.
Diagnosis of prediabetes and T2DM by glycemia ≥ 100 mg/dl and Hb A1c ≥ 5·7 %.
P-values for interaction between total dairy products intake and BMI for prediabetes and diabetes = 0·824.
* Proportion of adolescents with frequency of prediabetes and diabetes detected in the study.
† Model 1: adjusted for age, sex, region (North, Northeast, Midwest, Southeast and South), skin colour and type of school.
‡ Model 2: adjusted for model 1 plus breakfast, physical activity, energetic intake, food groups (fruits, vegetables, whole grains, beans and legumes, processed meat, snacks and sweets, sugary drinks), smoking and screen time.
§ Model 3: adjusted for model 2 plus BMI.
We observed that adolescents with higher consumption of low-fat dairy products had a 46 % (95 % CI 1·18, 1·80) higher prevalence for the outcome. In subgroup analysis, stratifying by BMI, we found that normal-weight adolescents with higher consumption of low-fat dairy products had 52 % (95 % CI 1·12, 2·05) higher PR of prediabetes and T2DM (Table 6).
Table 6. Poisson regression of the association between prediabetes and diabetes detected in the study and consumption category of dairy products subgroups (whole and low fat) and by BMI
PR, prevalence ratio.
Values are adjusted PR and 95 % CI.
Diagnosis of prediabetes and diabetes by glycemia ≥ 100 mg/dl and HbA1c ≥ 5·7 %.
P-values for interaction between dairy products subgroup and BMI for the outcome: whole dairy products = 0·926, low-fat dairy products = 0·448.
* ERICA sample low consumption: ≤ Median (P50).
† ERICA sample high consumption: > Median (P50).
‡ Proportion of adolescents with altered levels of prediabetes and diabetes markers detected in the study.
§ Model 1: adjusted for age, sex, region (North, Northeast, Midwest, Southeast and South), skin colour and type of school.
|| Model 2: adjusted for model 1 plus breakfast, physical activity, energetic intake, food groups (fruits, vegetables, whole grains, beans and legumes, processed meat, snacks and sweets, and sugary drinks), smoking and screen time.
¶ Model 3: adjusted for model 2 plus BMI.
In online Supplementary Table 3, we observed that the higher consumption of yogurt is inversely associated with fasting glucose levels −0·868 mg/dl (95 % CI −1·568, −0·167), and the higher consumption of butter is associated directly with HbA1c: 0·025 % (95 % CI 0·001, 0·048), after adjusting for confounding factors. When we stratified the sample by BMI, it was observed that higher milk consumption is inversely related to fasting blood glucose levels of −0·939 (95 % CI −1·862, −0·017) in the normal-weight sample. In the overweight sample, a direct association was found between the higher consumption of butter and HbA1c: 0·058 (95 % CI 0·012, 0·105; P = 0·014). It was also found that the higher consumption of cheese was associated with a higher combined PR of 1·33 (95 % CI 1·13, 1·56) for prediabetes and T2DM in the total sample. Similar results were found in the normal-weight sample (online Supplementary Table 4).
Discussion
In the present study, higher consumption of total dairy products was inversely associated with fasting blood glucose levels. However, when analysed by BMI categories, this association was only observed in those with overweight and obesity, including HOMA-IR levels. There was an inverse association between full-fat dairy products consumption and fasting blood glucose and HOMA-IR in overweight and obese adolescents and between higher yogurt consumption and fasting glucose levels. Finally, a lower combined prevalence of prediabetes and T2DM was detected only in overweight and obese adolescents that presented 2–3 servings per d of dairy products consumption, compared with less than 2 servings per d and, we found a higher prevalence of prediabetes and T2DM among those who reported higher consumption of low-fat dairy products and cheese in the total sample and among adolescents with normal BMI, respectively.
The average consumption of dairy products in ERICA was 2·3 servings per d, which is below what is recommended by the Dietary Guidelines for Americans(40), which recommends three cups of dairy products per d. In Europe, recommendations range from three to four servings or between 450 to 600 ml of dairy products per d, according to the National Dietary Guidelines for Greece(Reference Kastorini, Critselis and Zota41) and Belgium (Flemish Food-Based Guideline)(42), respectively. In Brazil, dairy products are mostly part of the group of natural or minimally processed foods that should be part of a healthy diet.
In ERICA, total dairy products consumption was inversely associated with fasting blood glucose levels. Data from the Nurses’ Health Study II, in a subsample of 37 038 women who completed a retrospective FFQ on their diet during adolescence, found that higher dairy products consumption (two servings per d) was associated with a lower risk of T2DM by 27 % (95 % CI 0·54, 0·97), after adjusting for risk factors in adolescence and adulthood, but the association was attenuated when adjusting for adult dairy products intake and weight changes since the age of 18 years(Reference Malik, Sun and van Dam43). Similar results in adults were observed in a systematic review, which mostly included studies from developed countries and reported a borderline reduction for T2DM (RR: 0·97; 95 % CI 0·95, 1·00) for each 200 g/d of total dairy products consumed(Reference Soedamah-Muthu and de Goede23). Results from a prospective cohort of 63 257 Chinese men and women showed that dairy food intake was significantly associated with a reduced risk of T2DM(Reference Talaei, Pan and Yuan44). In Brazil, a study with an adult population also found that total dairy products intake was inversely associated with fasting blood glucose (−0·46 ± 0·20 mg/dl), HbA1c (−0·02 ± 0·00 %) and HOMA-IR (−0·04 ± 0·02)(Reference Drehmer, Pereira and Schmidt34). However, the literature differs from these findings, as Chen et al.,(Reference Chen, Wang and Tong45) in a systematic review of fourteen studies, did not find a significant association between total dairy products consumption and T2DM. The association related to the benefits of full-fat dairy products may be due to its content of bioactive lipids such as trans palmitoleic acid (tC16:1) which regulates metabolic and inflammatory pathways(Reference Hirahatake, Slavin and Maki20) and is associated with lower insulin levels and incidence of T2DM(Reference Mozaffarian, de Oliveira Otto and Lemaitre46). Other dairy products compounds such as protein (80 % casein and 20 % whey) can increase satiety and control appetite, helping energy balance. Evidence also points to proteins increasing the secretion of gastrointestinal hormones and increasing diet-induced thermogenesis(Reference Bendtsen, Lorenzen and Bendsen47).
In the analyses stratified by BMI, the inverse association between total consumption of dairy products, especially full-fat dairy products, and fasting glucose levels and HOMA-IR was restricted to the sample of adolescents with excess weight (overweight or obesity). Similar results were reported by Pereira et al.,(Reference Pereira, Jacobs and Van Horn48) in a sample of four US metropolitan areas that included 3157 adults aged between 18 and 30 years and who were followed from 1985–1986 to 1995–1996, showing an inverse association between total (whole-fat, specifically) dairy products consumption and blood glucose among overweight and obese adults. A possible explanation for this association of dairy products and overweight or obese adolescents may be due to the ability of dairy products to significantly attenuate oxidative and inflammatory stress(Reference Stancliffe, Thorpe and Zemel49,Reference Zemel, Sun and Sobhani50) .
In our analysis, there was an inverse association between higher consumption of full-fat dairy products and fasting glucose and HOMA-IR only in the overweight sample. These data are consistent with the study by Hruby et al.,(Reference Hruby, Ma and Rogers51) which included 1867 prediabetes-free participants and concluded that consumption of full-fat dairy products was associated with a 25 % lower risk of prediabetes but also found that full-fat and low-fat dairy products had a 39 % and 32 % lower risk, respectively. In the present study, we showed that the higher consumption, compared with the lower consumption of low-fat dairy products, was associated with a 46 % and 51 % higher combined prevalence of prediabetes and T2DM in the total sample and in the normal-weight sample, respectively. However, there is a lack of consistency in the literature as different results were found in systematic reviews(Reference Gijsbers, Ding and Malik14,Reference Gao, Ning and Wang25,Reference Aune, Norat and Romundstad52) , carried out in developed countries, which concluded that low-fat dairy products had an inverse linear association with T2DM and found no association for full-fat dairy products. A possible explanation for the finding that points to a higher likelihood of T2DM, related to the consumption of low-fat dairy products, may be due to reverse causality, since the family history of T2DM or previous obesity itself influences the tendency to choose low-fat dairy products.
As for the analysis of the other dairy products subgroups, we found an inverse association between higher yogurt consumption and fasting blood glucose levels. Similar results were found in the systematic review(Reference Gijsbers, Ding and Malik14), which included twenty-two prospective cohort studies with a total of 579 832 adults, and which observed non-linear inverse associations for yogurt intake. Chen and colleagues(Reference Chen, Sun and Giovannucci53) found a 14 % reduction in the risk of T2DM (RR 0·86 95 % CI 0·83, 0·90; P < 0·001) for every 80 g of yogurt consumption per d. Higher consumption of cheese was associated with a higher combined prevalence of prediabetes and T2DM, results that only persisted in adolescents with normal BMI. A systematic review that included three large prospective cohorts of US men and women found similar results, with an increase in cheese consumption of more than half a serving a day associated with a 9 % higher risk of T2DM(Reference Drouin-Chartier, Li and Ardisson Korat54). Different results were found in a systematic review where the consumption of 50 g of cheese showed an inverse association with the risk of T2DM, 0·92 (95 % CI 0·86, 0·99)(Reference Aune, Norat and Romundstad52).
Dairy products have beneficial nutrients in their composition related to a lower risk of T2DM(Reference Guo, Givens and Astrup55). This association between dairy products and T2DM may be due to specific amino acids, medium- and long-chain saturated fats, unsaturated fats, branched-chain fats, probiotics, vitamin K1/K2, and Ca(Reference Mozaffarian and Wu56). A prospective cohort study found that dairy Ca was associated with a 16 % lower risk of T2DM (95 % CI 0·76, 0·93); however, no association was found for non-dairy Ca(Reference Talaei, Pan and Yuan44). Another important nutrient is milk protein, which due to its amino acids and bioactive peptides could improve the insulin response and lead to lower postprandial glucose levels(Reference Hidayat, Du and Shi57).
Liu et al.(Reference Liu, van der Schouw and Soedamah-Muthu58) investigated the association between dietary SFA and the risk of T2DM in a cohort of 37 421 participants over 10 years and found that milk, dairy products and butter-derived SFA were not associated with the risk of T2DM but found that cheese-derived SFA were associated with a lower risk of T2DM (HR 0·90 95 % CI 0·83, 0·98). These findings, contrary to the present study, may be due to the type of cheese produced in the Netherlands and the use of this country’s composition table.
The present study has some limitations. Although ERICA is a study with a large sample size, the response rate for the completion of all procedures was 52 %, which may influence the results(Reference da Silva, Klein and Souza59). As it is a cross-sectional study, causal inferences are limited and residual confounding is possible. It is noteworthy that a single 24-h dietary recall was utilised for most of the sample, which may not accurately reflect their typical diet(Reference Bloch, Szklo and Kuschnir27). To mitigate this limitation, a second recall was randomly administered to 10 % of the sample. However, it should be noted that this approach may not be adequate to correct for intrapersonal error and provide an accurate estimate of habitual food consumption(Reference Tooze, Kipnis and Buckman60,Reference Willett61) . Data are also limited on the details of certain types of dairy products and their fat and nutrient profiles. Blood results were analysed using standardised procedures at a central laboratory. But it is important to highlight that ERICA is the largest study with a representative sample of Brazilian adolescents with detailed measurements of food consumption and blood measures.
There are few studies analysing the association between dairy products and T2DM in adolescents. The HELENA study (with a sample of adolescents aged between 12·5 and 17·5 years, from eight European cities) found that dairy products are the food group that best identifies adolescents with lower cardiovascular risk(Reference Santaliestra-Pasías, Bel-Serrat and Moreno62). In Brazil, the latest surveys of Brazilian family budgets point to a decrease in the consumption of dairy products and their replacement with sweetened beverages in the general(63). The findings of the present study may help to reinforce the recommendations for dairy products consumption in the young population.
In summary, the present study showed that a higher intake of total dairy products was associated with lower blood glucose levels, specifically full-fat dairy products. In overweight adolescents, the results were more evident. It was also observed that low-fat dairy products were associated with a higher combined prevalence of prediabetes and T2DM in the total sample. Longitudinal studies are needed in the adolescent population to increase understanding of the role of dairy product intake in the aetiology of T2DM early in life.
Acknowledgements
This work was supported by the Innovation and Research Funding (FINEP; grant: 01090421); Brazilian National Research Council (CNPq; grants: 565037/2010-2, 405009/2012-7 and 457050/2013-6); Research and Events Incentive Fund (FIPE) of the Hospital de Clínicas de Porto Alegre (2009-0098). This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES) - Finance Code 001.
M. M., F. V. C. and M. D. contributed to concept development and interpreted the data. V. B. L., M. M. and F. V. C. conducted the analysis. M. M. and P. B. Z. R. were responsible for execution of the study and processing the data. M. M. drafted the manuscript. B. D. S., M. A. P., F. V. C. and M. D. critically reviewed the manuscript. All authors have responsibility for the final content.
The authors declare that there are no conflicts of interest.
Supplementary material
For supplementary material/s referred to in this article, please visit https://doi.org/10.1017/S0007114523001356
