Depression is a chronic mental illness with great worldwide prevalence(1). Data indicate that 3·8 % of the world population is affected by depression, affecting about 5 % of the adult population and 5·7 % of the population over 60 years old(1). The main characteristics of depressive disorder are depressed mood, feelings of guilt or low self-esteem, changes in sleep and appetite, lack of disposition and poor concentration. In many cases, these symptoms can be accompanied by feelings of anxiety(Reference Marcus, Yasamy and van Ommeren2). In moderate or severe intensities, depression can greatly affect an individual’s daily life in work, studies and family relationships, and in its most severe forms, it can lead to suicide(1,Reference Marcus, Yasamy and van Ommeren2) .
In this sense, much has been investigated about the pathogenesis of depression, and oxidative stress has stood out among the risk factors(Reference Ng, Berk and Dean3,Reference Visentin, Colombo and Scotton4) . Oxidative stress can be characterised as an imbalance between the antioxidant defences of the organism and the presence of free radicals, a pathogenic process related to cell injury and death(Reference Ng, Berk and Dean3–Reference Kruk, Aboul-Enein and Kładna5). It is noteworthy that the brain, compared with other organs, is highly vulnerable to oxidative stress due to its high metabolism(Reference Kruk, Aboul-Enein and Kładna5,Reference Lehtinen and Bonni6) . Thus, redox imbalance may be related to depression through mechanisms such as inflammation and neurodegeneration, impairing neuronal and neurotransmitter function(Reference Ng, Berk and Dean3,Reference Visentin, Colombo and Scotton4,Reference Bajpai, Verma and Srivastava7) .
Due to the need for new approaches to treat and prevent depression, antioxidants such as vitamins C, E and Zn have been associated with improvements in neurocognitive function, bringing therapeutic benefits to depression(Reference Du, Zhu and Bao8). A recent cross-sectional study with 14 737 individuals participating in the Brazilian Longitudinal Study of Adult Health (ELSA) reported, for Brazilian women, an inverse association of the consumption of Zn, Se, vitamin A and C with depression(Reference Ferriani, Silva and Molina9). Inverse associations of Zn and Se consumption with depression were also found for 14 834 adults participating in the National Health and Nutrition Examination Survey 2009–2014(Reference Li, Wang and Xin10).
Despite the results obtained so far between nutrient intake and the occurrence of depression, the isolated assessment of antioxidants may not be as effective as the assessment of the interaction of different dietary antioxidants(Reference Serafini and Del Rio11). Therefore, the Dietary Total Antioxidant Capacity (dTAC), an index capable of measuring the global content of dietary antioxidants(Reference Carlsen, Halvorsen and Holte12), has proved to be a helpful tool for investigating the interaction between dietary antioxidants and health outcomes(Reference Sabião, Bressan and Pimenta13–Reference Hermsdorff, Puchau and Volp15). However, investigations into the relationship between dTAC and depression are still limited, mostly being cross-sectional studies with the Iranian population(Reference Pereira, Da Silva and Hermana16–Reference Milajerdi, Keshteli and Afshar19). Unfortunately, as far as we know, no prospective study has investigated the association between dTAC and depression in Brazilians, nor has it analysed this relationship through the Total Antioxidant Capacity of food groups (fgTAC)(Reference Pereira, Da Silva and Hermana16). Thus, the present study aimed to assess the association between dTAC and fgTAC with the incidence of depression in Brazilian graduates participating in the Cohort of Universities of Minas Gerais (CUME Study).
Methodology
Cohort of Universities of Minas Gerais (CUME Study)
The CUME Study is an open, prospective cohort conducted with alumni from universities located in the state of Minas Gerais (Brazil), whose main objective is to assess the impact of the Brazilian dietary pattern, specific diet factors and the nutritional transition in the incidence of chronic non-communicable diseases, as previously detailed(Reference Domingos, Da Silva Miranda and Pimenta20).
The CUME Study questionnaires were developed on the Alchemer (www.alchemer.com) online interface by specialists. The team conducted pilot studies with the baseline questionnaire to assess its applicability(Reference Domingos, Da Silva Miranda and Pimenta20). The CUME Study began in 2016 with the application of the baseline questionnaire Q_0, and its recruitment has been periodic since then. Invitations to participate were sent to all volunteers who had emails available. Thus, every 2 years, the participants are invited to answer follow-up questionnaires (Q_2, Q_4…Q_n) to update their information about lifestyle, the emergence of new diseases and changes in dietary patterns, among others, while new potential participants are recruited and invited to answer the baseline cohort questionnaire (Q_0). Characteristics related to project design and recruitment of the first volunteers were described in a previous study(Reference Domingos, Da Silva Miranda and Pimenta20).
The questionnaires are answered in a virtual environment of the CUME Study. Q_0 is divided into two phases: the first phase with questions about socio-demographic characteristics, lifestyle, biochemical markers (TAG concentrations, total cholesterol, HDL-cholesterol, LDL-cholesterol, blood glucose concentration) and related to the individual’s health outcomes; the second phase that is answered after 1 week contains a FFQ and questions related to dietary practices. On the other hand, Q_2 is composed of questions related to changes in lifestyle, eating habits and health conditions. Finally, the follow-up questionnaire Q_4 has questions related to socio-demographic characteristics, lifestyle, biochemical markers, health outcomes and insomnia severity.
This study was conducted according to the guidelines laid down in the Declaration of Helsinki, and all procedures involving human subjects were approved by the Human Research Ethics Committees of all participating institutions: Federal University of Minas Gerais (CAAE registration number: 07223812.3.3001.5153); Federal University of Viçosa (CAAE registration number: 4483415.5.1001.5149); Federal University of Ouro Preto (CAAE registration number: 44483415.5.2003.5150); Federal University of Lavras (CAAE registration number: 44483415.5.2002.5148); Federal University of Juiz de Fora (CAAE registration number: 4483415.5.5133); Federal University of Vale do Jequitinhonha e Mucuri (CAAE registration number: 44483415.5.2005.5103) and Federal University of Alfenas (CAAE registration number: 4.501.344). Written informed consent was obtained from all subjects(Reference Domingos, Da Silva Miranda and Pimenta20).
Data collection
Information from the Q_0 questionnaires for the years 2016, 2018 and 2020 composed the baseline of the present study. We used data from the 2-year follow-up Q_2 and the 4-year follow-up Q_4 to investigate the incidence of the depression outcome.
The first collection related to the 2-year follow-up occurred in 2018, answered by participants who started the baseline (Q_0) in 2016. The second collection, associated with Q_2, occurred in 2020 when we invited participants who completed the Q_0 follow-up questionnaire in 2018. Finally, the collection related to the Q_4 follow-up questionnaire occurred in 2020. Participants who answered Q_2 in 2018 were invited to fill it out.
Study population
The CUME Study has responses from 7710 participants. For the present study, of the total number of respondents, we excluded 4084 participants who had not accomplished at least the 2-year follow-up. We also excluded women who were pregnant or had been pregnant in the last year (n 430), participants with energy consumption < 2092 or > 25104 kJ per day (< 500 kcal or > 6000 kcal) (n 79)(Reference Schmidt, Duncan and Mill21), participants who reported not having lived in Brazil in the last year (n 135) and foreigners living in Brazil (n 10) and participants who reported a diagnosis of depression at baseline (n 400). The final sample consisted of 2572 participants who answered at least one follow-up questionnaire (Fig. 1).
Fig. 1. Flow chart of participant selection.
Outcome variable: incidence of depression
We considered incident cases of depression in those participants who were disease-free at the beginning of the follow-up and were classified as having the disease during Q_2 or Q_4 follow-up. For it, we considered incident depression when participants answered yes to the following question: ‘Since the previous questionnaire, were you clinically diagnosed with depression for the first time?’. Individuals who reported using antidepressants but did not confirm the diagnosis of depression were not included due to the possibility of the therapeutic use of such drugs in diseases other than depression. The reliability between medical diagnosis and self-report depression in this cohort was validated in a subsample of participants showing good agreement (81·0 %) with a κ value of 0·62(Reference Santos, Oliveira and Miranda22).
Food consumption and estimation of Dietary Total Antioxidant Capacity
We assessed habitual food consumption using FFQ, previously validated for the study population(Reference Azarias, Marques-Rocha and Miranda23), which includes 144 food items presented in the following groups: dairy products, meat and fish, cereals and legumes, fruits, vegetables, fats, and oils, beverages and other foods. Each participant reported the frequency of consumption of a particular food (daily, weekly, monthly or yearly), the number of times they consumed it (0–9 or more times) and the portion size. To facilitate filling in the portion sizes of food items and obtain the most reliable information possible, participants had access, at the time of answering the FFQ, to images of portions of food and serving utensils from the photo album with ninety-six pictures prepared by the CUME Study team(Reference Domingos, Da Silva Miranda and Pimenta20). The consumption of macro and micronutrients was calculated using primarily data from the Table of Nutritional Composition of Foods consumed in Brazil(24) and, in the absence of information in this table, we consulted the Brazilian Table of Food Composition(25) and the USDA National Nutrient Database(Reference Gebhardt and Thomas26).
We used the values from the Ferric Reducing Antioxidant Power (FRAP) assay to estimate dTAC, which measures the antioxidant capacity of food in the presence of Fe. We consulted previously published databases to obtain the FRAP values(Reference Carlsen, Halvorsen and Holte12,Reference Koehnlein, Bracht and Nishida27) . Thus, the dTAC of each food item resulted from the multiplication between the amount in grams of food consumed and its corresponding FRAP value in mmol per gram of food (mmol/g).
We used the following criteria to assign the FRAP value to a portion of food: when there was more than one FRAP analysis value for the same food, we considered the mean value; in the absence of a FRAP value for a particular food, when possible, we used the value of a similar food from the same botanical group, or the same food in a different way of preparation. We did not assign FRAP values to foods where there was no record of this value, and it was not possible to estimate for foods from similar botanical groups or different methods of preparation
A total of 133 food items were covered with FRAP values. We summed all FRAP values of foods reported in the FFQ to estimate the total dTAC of each participant. In order to perform a sensitivity analysis, we also calculated dTAC values, excluding coffee values, as this is a beverage that greatly contributes to dTAC values in our population(Reference Sabião, Bressan and Pimenta13). We also calculated fgTAC values according to food groups (fruits, vegetables, beans and lentils, oilseeds, dairy products, meats and eggs, pasta, breads and cereals, oils and fats, junk food, natural juices, teas and coffees, artificial juices and sodas, and alcoholic beverages) (online Supplementary Table 1).
We adjusted all food consumption variables for daily energetic intake using the residual method(Reference Willett28), including dTAC.
Covariates
We obtained the other variables from self-reported information in the baseline questionnaire Q_0, including socio-demographic variables (sex, skin colour, marital status, professional status), use of vitamin supplements and smoking habits (non-smoker, smoker or ex-smoker). The frequency of heavy episodic consumption of alcoholic beverages (1–2 d/month, 3–4 d/month and 5 or more days/month) was also a variable included in this study, with heavy episodic consumption considered as 4 or more doses of alcoholic beverage on a single occasion for women and 5 or more doses of alcoholic beverage on a single occasion for men(29). We also assessed physical activity using a list of twenty-four activities expressed in minutes per week. Participants with ≥ 150 min/week of moderate-intensity activity or ≥ 75 min/week of vigorous-intensity activity were considered active; participants with < 150 min/week of moderate-intensity activity or < 75 min/week of vigorous-intensity activity were classified as insufficiently active and those who reported no leisure-time physical activity were classified as inactive(30). We obtained the BMI from the self-reported weight and height in the baseline questionnaire, dividing weight (kg) by height (m) squared. BMI values ≥ 30 kg/m2 were classified as obesity(31,32) . The BMI based on self-reported weight and height was also validated, in a previous study, for the population participating in the CUME Project, indicating excellent agreement with the measured intra-class correlation coefficient 0·989(Reference Miranda, Ferreira and Oliveira33).
Statistical analysis
We performed the analyses with Stata SE 15·0. A two-tailed P-value less than 0·05 was considered statistically significant.
We described socio-demographic, lifestyle, health and food consumption characteristics in absolute or relative frequency or as mean and standard deviation according to the dTAC quartiles baseline. We calculated the P-value using Pearson’s chi-squared tests for categorical variables and ANOVA for continuous variables to compare the categories.
The follow-up time was calculated in person-years for each participant: difference between the date of completion of the follow-up questionnaire in which depression was diagnosed and the date of completion of the baseline questionnaire. In this sense, we created Cox regression models to assess the association between dTAC in quartiles and the incidence of depression. We used the lowest quartile as the reference category to compare the incidence of depression among the dTAC categories (exposure variable). As a guide for selecting the covariates included in the analyses, we constructed a directed acyclic graph (DAG) using the DAGitty program. DAG are a strategy that helps identify a minimum set of confounding covariates in the analysis of causal relationships, helping to estimate less biased measures of effect(Reference Textor and Hardt34,Reference Greenland, Pearl and Robins35) (online Supplementary Fig. 1). Thus, we adjusted the final model for potential confounders such as sex, age (continuous), smoking habit (non-smoker/smoker/ex-smoker), frequency of heavy episodic alcohol consumption (never/from 1 to 2 days per month/from 3 to 4 days per month/5 or more days per month), marital status (single/married or stable union/separated or divorced or widowed), skin colour (white/non-white), professional status, physical activity (inactive/insufficiently active/active), use of vitamin supplements (yes/no), BMI (continuous) and energetic intake (kcal/d) and vitamin D consumption (μg/d). Linear trends were tested using the median values of each quartile of the exposure variable ordered in Cox regression models. We also performed a sensitivity analysis excluding coffee items from the dTAC computation. In addition, we analysed the relationship between fgTAC and the incidence of depression (fruits, vegetables, beans and lentils, oilseeds, dairy products, meat and eggs, breads, pasta and cereals, oils and fats, ‘junk food’, natural juices, teas and coffees, artificial juices and soda and alcoholic beverages), considering that antioxidants present in diet may be inserted in different proportions in food matrices(Reference Hermsdorff, Puchau and Volp15,Reference Hermsdorff, Zulet and Puchau36) .
Results
During the average time of 2·96 (1·00) years of monitoring of the present study, 246 new cases of depression (32·3/1000 person-years) were observed and the mean age of the participants was 36·09 (SD 9·63). The mean dTAC was 11·03 (4·84 mmol/d).
Participants included in the highest quartile for dTAC (> 13·32 mmol/d) are mostly older, married or in a stable union, employed workers, smokers and physically active, in addition to having a higher frequency of heavy episodic alcohol consumption (Table 1). Regarding food consumption, individuals belonging to the fourth quartile of dTAC, when compared with the first quartile (< 7·92 mmol/d), had a higher intake of carbohydrates, n-3, alcohol, vitamins A, E, C and B9, Mg, fibre, as well as higher consumption of fruits and vegetables (Table 2).
Table 1. Baseline socio-demographic and health characteristics according to energy-adjusted dTAC (mmol/d) quartiles, CUME Project
(n 2572) (Mean values and standard deviations; numbers and percentages)
dTAC, Dietary Total Antioxidant Capacity.
P values according to trend chi-squared test.
Table 2. Baseline dietary intake according to the energy-adjusted dTAC (mmol/d) quartiles, CUME Project (n 2572)
(Mean values and standard deviations)
Data expressed as mean (standard deviation).
* P values by ANOVA test. Different letters show statistically significant differences between groups according to Bonferroni’s post hoc test.
There was no association between dTAC and incidence of depression (Table 3), regardless of adjustment for confounding variables. The results of the analysis of dTAC without coffee and incidence of depression remained similar. When we assessed the associations between fgTAC and the incidence of depression (Tables 4 and 5), we observed a lower incidence of depression according to the quartiles of fgTAC of natural juices in the model adjusted by age and sex (hazard ratio: 0·70; 95 % CI 0·49, 0·99). Still, the significance of this association was lost after the total adjustment of the model. On the other hand, the TAC of the beans and lentils was inversely associated with the incidence of depression in our cohort (hazard ratio: 0·61; 95 % CI 0·41, 0·90) (Table 4). Interestingly, a higher fgTAC from the junk food group was positively associated with a higher incidence of depression among participants after all adjustments (hazard ratio: 1·57; 95 % CI 1·08, 2·26).
Table 3. Hazard ratios and 95 % CI of depression incidence according to dTAC and dTAC without coffee, CUME Project
(n 2572) (Mean values and standard deviations)
dTAC, Dietary Total Antioxidant Capacity.
* Adjusted – Sex and age.
† Adjusted – Model 1 + smoking status (never, current, former), alcohol consumption (BINGE frequency) and vitamin D consumption (mcg).
‡ Adjusted – Model 2 + marital status (single/married or stable union/separated or divorced or widowed), skin colour (white and not white), physical activity (inactive/insufficiently active/active), use of supplements (yes or no), energy intake (continuous, kcal/d), baseline BMI (continuous kg/m2), professional situation.
Table 4. Hazard ratios and 95 % CI of depression incidence according to fgTAC from food groups, CUME Project
(n 2572) (Mean values and standard deviations)
fgTAC, Total Antioxidant Capacity of food groups.
* Adjusted – Sex and age.
† Adjusted - Model 1 + smoking status (never, current, former), alcohol consumption (BINGE frequency) and vitamin D consumption (mcg).
‡ Adjusted – Model 2 + marital status (single/married or stable union/separated or divorced or widowed), skin colour (white and not white), physical activity (inactive/insufficiently active/active ), use of supplements (yes or no), energy intake (continuous, kcal/d), baseline BMI (continuous kg/m²), professional situation.
Table 5. Hazard ratios and 95 % CI of depression incidence according to consumption of fgTAC by beverages, CUME Project
(n 2572) (Mean values and standard deviations)
fgTAC, Total Antioxidant Capacity of food groups.
* Adjusted – Sex and age.
† Adjusted – Model 1 + smoking status (never, current, former), alcohol consumption (BINGE frequency) and vitamin D consumption (mcg).
‡ Adjusted – Model 2 + marital status (single/married or stable union/separated or divorced or widowedd), skin colour (white and not white), physical activity (inactive/insufficiently active/active), use of supplements (yes or no), energy intake (continuous, kcal/d), baseline BMI (continuous kg/m²), professional situation.
Discussion
In the present study, we found no association between total dTAC and incidence of depression. Still, fgTAC from specific food groups, such as natural juice, beans and lentils, and junk foods, showed associations. As far as we know, this is the first prospective study to investigate the association between dTAC and fgTAC by different food groups with the incidence of depression in Brazilian graduates.
When analysing the fgTAC, we observed an inverse association between the fgTAC of the beans and lentils group with the incidence of depression. A study from the National Health Survey in Brazil with 46 785 adults showed an inverse association between bean consumption and depression(Reference Sousa, Marques and Levy37). Bean consumption was also inversely associated with mental disorders in a study with 712 Brazilian pregnant women(Reference Paskulin, Drehmer and Olinto38). It is noteworthy that beans, besides being sources of antioxidants such as polyphenols, are sources of B vitamins and minerals such as Fe, K and Mg, in addition to dietary fibre(Reference Silva, Rocha and Brazaca39,Reference Velásquez-Meléndez, Mendes and Pessoa40) . In a previous study, beans proved to be an important contributor to folate consumption in part of the CUME Project baseline population(Reference Pereira, Bressan and Oliveira41). In fact, in addition to antioxidants, folate consumption has been inversely associated with depression(Reference Miyaki, Song and Htun42,Reference Murakami, Miyake and Sasaki43) . Another point worth mentioning is the fact that the regular consumption of beans may be related to a higher quality dietary pattern, characterised by a diversity of in natura and minimally processed foods, while low intake may be associated with an increase in the consumption of ultra-processed foods(Reference Velásquez-Meléndez, Mendes and Pessoa40,Reference Rodrigues, da Costa Proença and Calvo44,Reference Silveira, de Novaes and Vieira45) . We emphasise that the dietary pattern rich in ultra-processed foods is positively related to the incidence of depression(Reference Sousa, Marques and Levy37,Reference Adjibade, Julia and Allès46) .
We observed that higher fgTAC of junk food was positively associated with higher incidence of depression in our cohort. This fact raises the question of the importance of the food matrix in which antioxidants are inserted. Although some ultra-processed foods have vitamins and minerals with antioxidant potential added to their composition to increase their shelf life, these are generally rich in simple sugars, fats, flavourings and preservatives that can contribute to a pro-oxidant and inflammatory state, closely related to depression(Reference Lobo and Tramonte47–Reference Contreras-Rodriguez, Solanas and Escorihuela49). In addition, a diet rich in fast food can contain a lower amount of vitamins and minerals than in natura or minimally processed foods. The deficient consumption of several nutrients is related to depression(Reference Louzada, Martins and Canella50,Reference Wang, Um and Dickerman51) .
We did not observe any association between total dTAC and incidence of depression. These findings agree with a prospective study with 911 Japanese workers, with no associations between dTAC and incidence of depressive symptoms after 3 years of monitoring(Reference Miki, Eguchi and Kochi52). In the same way, two cross-sectional articles, one with climacteric women and another with sixty Iranian men, found no association between dTAC, depressive symptoms or diagnosed depression(Reference De Oliveira, Teixeira and Theodoro53,Reference Prohan, Amani and Nematpour54) . Contrary to our findings, three cross-sectional Iranian studies observed positive associations between dTAC and the prevalence of depressive symptoms(Reference Abshirini, Siassi and Koohdani17–Reference Milajerdi, Keshteli and Afshar19). In a recent systematic review, our group analysed existing studies that linked dTAC and depression, concluding that consumption of an antioxidant-rich diet characterised by high dTAC scores appears to be inversely associated with depression, anxiety and sleep disorders. However, we emphasise that there are few studies available in the literature, and most have a cross-sectional design and methodological limitations, as they were conducted with Iranian individuals and, most of them, with women(Reference Pereira, Da Silva and Hermana16).
Contrary to our expectations, we did not find associations between the fgTAC of the fruit and vegetables group and the incidence of depression. On the other hand, for the fgTAC of the natural juices group, the inverse association with depression remained only for the first adjustments (sex, age), not being maintained for total adjustments. A longitudinal study with Add Health Study data, which monitored 3696 17-year-old participants for 12 years, found no association between fruit and vegetables consumption and the incidence of depression either(Reference Hoare, Hockey and Ruusunen55). In turn, another longitudinal study with 8353 Canadians that observed inverse associations between fruit and vegetable consumption and depression being attenuated after adjusting variables such as smoking and physical activity(Reference Kingsbury, Dupuis and Jacka56). Although the results of the consumption of fruits and vegetables are contradictory, a meta-analysis with observational studies found a reduction in the risk of depression with the increase in the consumption of fruits and vegetables(Reference Opie, Itsiopoulos and Parletta57). It is worth mentioning that fruits and vegetables and natural juices are sources of antioxidants that modulate oxidative stress, and their consumption is related to mental health(Reference Kingsbury, Dupuis and Jacka56,Reference Opie, Itsiopoulos and Parletta57) . In addition, they are related to healthier eating patterns(Reference Silveira, de Novaes and Vieira45). However, the association between the consumption of fruits and vegetables and other behavioural factors in the incidence of depression can be complex(Reference Kingsbury, Dupuis and Jacka56). Thus, some behavioural factors such as physical activity, use of supplements, BMI and professional situation may have a more important impact on depression when compared with, fruits, vegetables and natural juices consumption.
The strengths of this study are its prospective design and the use of the quantitative FFQ previously validated for the study population, with good validity and reproducibility, ensuring good consistency in food consumption analyses(Reference Azarias, Marques-Rocha and Miranda23). In addition, we highlight that the self-report of depression was previously validated for the study population(Reference Santos, Oliveira and Miranda22). Another point is the high level of education of the participants, which can result in more reliable answers and greater adherence to the study(Reference Seguí-Gómez, de la Fuente and Vázquez58). Finally, we highlight the use of several confounding factors for our adjustments, carefully chosen after literature reviews and with the help of a directed acyclic graph(Reference Textor and Hardt34,Reference Greenland, Pearl and Robins35) .
As limitations, we highlight that although the FFQ has good reproducibility, we cannot guarantee that the baseline FRAP values represent the habitual long-term dietary intake precisely. Another point is the lack of national tables for FRAP values, requiring the use of values arranged in international tables for most calculations. Such factors may mediate the results observed here. Nor can we discard the possibility of residual confounding by some unmeasured or not precisely measured factors. We highlight the non-assessment of plasma TAC, but it is worth mentioning that plasma TAC may not be reflected in long-term diets, which limits its comparison with dTAC(Reference Nascimento-Souza, Paiva and Martino14,Reference Pellegrini, Vitaglione and Granato59) . In addition, dTAC proves to be a handy tool in assessing the relationship between diet and health outcomes(Reference Nascimento-Souza, Paiva and Martino14,Reference Pellegrini, Vitaglione and Granato59) . Finally, although the collection of data from the follow-up questionnaire Q_4 was carried out in the initial months of the COVID 19 pandemic, we cannot guarantee that this short period of time had an influence on the increase in medical diagnosis of depression. However, future analyses carried out with the next years of follow-up of the cohort may provide answers about the impact of the pandemic on the incidence of depression.
Conclusion
Our findings do not support an association between dTAC and the incidence of depression after an average of 2·96 (1·00) years of follow-up in a highly educated Brazilian population. However, the inverse association of fgTAC from beans and lentils and the direct association of junk food with the incidence of depression in the population indicate that not only the presence of antioxidants but also the food matrix in which these antioxidants are inserted should be considered to explain the associations between diet and health outcomes. We highlight the need for further prospective studies with different nationalities to confirm these results.
Acknowledgements
The authors wish to express our gratitude to all the participants of The Cohort of Universities of Minas Gerais. The authors thank CAPES (Ministry of Education, Brazil) for granting a doctoral fellowship to G.A.P.S and A.C.G.L. J. B., H. H. M. H. and A. M. P. are research productivity fellows of CNPq (Ministry of Science and Technology, Brazil).
This study was financed in part by the Coordenação de Aperfeiçoamento de Pessoal de Nível Superior – Brasil (CAPES) – Finance Code 001. Minas Gerais Research Foundation (FAPEMIG) for the grants has supported the CUME Project (CDS-APQ-00571/13, CDS-APQ-02407/16 and CDS-APQ-00424/17).
Conceptualisation, methodology, formal analysis, investigation, data curation and writing, G. A. P. S., H. H. M. H., A. C. G. L., A. M. P., J. B., A. P. B. M. and A. S. A. All authors read and approved the final manuscript. All authors critically reviewed the manuscript and approved the final version submitted for publication.
The authors declare no conflict of interest.
Supplementary material
For supplementary materials referred to in this article, please visit https://doi.org/10.1017/S0007114523000181
