Methods

Study population

The ‘Seguimiento Universidad de Navarra’ (SUN) Project is a multipurpose Spanish cohort composed of university graduates. Baseline assessment and follow-up information was gathered through postal or web-based questionnaires collected every 2 years. The recruitment of participants started on 21 December 1999. It is a dynamic cohort – that is, recruitment is continuously open. The overall retention in the cohort approaches 90 %. Further details about the methodology and characteristics of the participants can be found in previously published reports⁽ Reference Segui-Gomez, de la Fuente and Vazquez ^{14 )}. The study was approved by the Institutional Review Board of the University of Navarra. Voluntary completion of the first questionnaire was considered to imply informed consent.

Through June 2014, 22 045 subjects had completed the baseline questionnaire of the SUN project. Subjects who had not completed at least one follow-up questionnaire, who were lost to follow-up, who were outside the pre-defined limits for energy intake (<3347·2 kJ/d (<800 kcal/d) or >16 736 kJ/d (>4000 kcal/d) in men and <2092 kJ (<500 kcal) or 14 644 kJ/d (>3500 kcal/d) in women⁽ Reference Willett ^{15 )}) and those who used antidepressant medication or had reported a previous clinical diagnosis of depression (lifetime prevalence) at baseline, or without date of diagnosis of incident depression (n 60), were excluded from the analyses. After these exclusions, 15 093 participants were finally included in this study (Fig. 1).

Fig. 1 Flow chart of participants. The Seguimiento Universidad de Navarra Project.

Exposure assessment

Dietary intake was assessed at baseline and after 10 years of follow-up with a validated semi-quantitative FFQ⁽ Reference Fernandez-Ballart, Pinol and Zazpe ^{16 ,} Reference De la Fuente-Arrillaga, Vázquez Ruiz and Bes-Rastrollo ^{17 )}. Nutrient intakes of 136 food items were calculated as frequency multiplied by the nutrient composition of a specified portion size for each food item, using an ad hoc computer programme specifically developed for this aim. A trained dietitian updated the nutrient database using the latest available information included in the food composition tables for Spain.

The dietary inflammatory index

The design and development of the DII has been described elsewhere⁽ Reference Shivappa, Steck and Hurley ^{18 )}. In brief, the DII is a scoring algorithm based on an extensive review of the literature published from 1950 to 2010, linking 1943 articles focussing on the effect of diet on six inflammatory biomarkers (IL-1β, IL-4, IL-6, IL-10, TNF-α and C-reactive protein). A total of forty-five food parameters including various macronutrients, micronutrients, flavonoids and individual food items linked to these inflammatory biomarkers were scored according to whether they increased (+1), decreased (−1) or had no effect (0) on inflammation. An overall food parameter-specific inflammatory effect score was calculated for each food item. This percentile was calculated by first linking the dietary data from a study to the regionally representative world database intake, which is based on actual human consumption in eleven populations from different parts of the world that provided a robust estimate of a mean and standard deviation for each parameter. These then become the multipliers to express an individual’s exposure relative to the ‘standard global mean’ as a Z-score. This is achieved by subtracting the ‘standard global mean’ from the amount reported and dividing this value by the standard deviation. To minimise the effect of ‘right skewing’, this value is then converted to a centred percentile score. The centred percentile score for each food parameter for each individual was then multiplied by the respective food parameter effect score, which is derived from the literature review, in order to obtain a food parameter-specific DII score for an individual. All the food parameter-specific DII scores were then summed to create the overall DII score for every participant in the study. The greater the DII score, the more pro-inflammatory the diet. More negative values represent more anti-inflammatory diets. The DII score could take on values ranging from about +8 (maximally pro-inflammatory) to −9 (maximally anti-inflammatory).

Construct validation of the DII was performed using data derived from two different sources of dietary intake information and serum high-sensitivity C-reactive protein as the construct validator⁽ Reference Shivappa, Steck and Hurley ^{18 )}. Thus far, the DII has been found to be associated with inflammatory cytokines including C-reactive protein and IL-6⁽ Reference Shivappa, Steck and Hurley ^{19 –} Reference Wood, Shivappa and Berthon ^{21 )}, the glucose intolerance component of MetS⁽ Reference Wirth, Burch and Shivappa ^{20 )}, increased odds of asthma and reduced forced expiratory volume in 1 s in an Australian population⁽ Reference Wood, Shivappa and Berthon ^{21 )}, shift work⁽ Reference Wirth, Burch and Shivappa ^{22 )}, colorectal cancer among women in the Iowa Women’s Health Study⁽ Reference Shivappa, Prizment and Blair ^{23 )}, prostate cancer⁽ Reference Shivappa, Bosetti and Zucchetto ^{24 )} and pancreatic cancer⁽ Reference Shivappa, Bosetti and Zucchetto ^{25 )}.

In this study, a total of twenty-eight food parameters were available from the FFQ, and therefore could be used to calculate DII (energy, carbohydrate, protein, total fat, alcohol, fibre, cholesterol, SFA, MUFA, PUFA, n-3, n-6, trans-fat, niacin, thiamin, riboflavin, vitamin B₁₂, vitamin B₆, Fe, Mg, Se, Zn, vitamin A, vitamin C, vitamin D, vitamin E, folic acid and caffeine.) The literature-derived inflammatory effect scores assigned to each of the food parameters are shown in Table 1.

Table 1 Scoring for each food parameter used for dietary inflammatory index calculation

* A negative value indicates anti-inflammatory effect and a positive score indicates pro-inflammatory effect.

Results

We recorded 1051 incident cases of depression during a median follow-up time of 8·5 years. Table 2 shows the distribution of the baseline characteristics of the participants according to the baseline DII categorised in quintiles. Participants with the highest DII were more likely to be men, single and younger and showed lower prevalence of CVD, DM2 or dyslipidaemia and lower daily energy intake. Some unhealthy lifestyle characteristics such as smoking behaviour and physical inactivity during leisure time were also more prevalent among participants with a more pro-inflammatory diet.

Table 2 Characteristics of participants according to quintiles of the dietary inflammatory index (Mean values and standard deviations; percentages)

DM2, type 2 diabetes; HTA, hypertension; METs, metabolic equivalents.

* Use of at least one of the following vitamin supplements: A, B₁, B₂, B₃, B₆, folic acid, B₁₂, C, D or E.

† A total of 8847 women were included.

The association between the DII and the risk of depression is shown in Table 3. A higher risk was found for the highest DII in the three models. In model 3 (adjusted for lifestyle factors and the presence of chronic diseases), a higher DII scores (fourth and fifth quintiles) compared with the lowest quintile of DII was associated with an approximately 25–35 % higher risk of depression (multiple-adjusted HR 1·24; 95 % CI 1·00, 1·53 for the fourth quintile and HR 1·37; 95 % CI 1·09, 1·73 for the fifth quintile). Moreover, a significant dose–response relationship was found (P _trend=0·015). When DII was updated using the information collected after 10 years of follow-up, the magnitude of the association was even higher. A relative increment in the risk of depression of approximately 50 % for the comparison between extreme quintiles of DII (HR 1·47; 95 % CI 1·17, 1·85; P _trend=0·010) was found. Additional adjustment for personal cancer history, marital status, alcohol intake and unemployment in the overall sample or for menopause status within women did not change the reported associations (data not shown).

Table 3 Risk of incident depression according to the adherence to quintiles of the dietary inflammatory index (DII) (Hazard ratios (HR) and 95 % confidence intervals)

Ref., referent value.

* Crude rates and 95 % confidence intervals.

† Model 1: adjusted for age and sex. Repeated measures. Cumulative average for DII (at baseline and after 10 years of follow-up). Energy intake, BMI and prevalence of diseases were also updated.

‡ Model 2: this includes all variables from model 1 plus BMI, smoking, physical activity during leisure time, use of vitamin supplements and total energy intake.

§ Model 3: this includes all variables from model 2 with additional adjustment for the presence of CVD, type 2 diabetes, hypertension and dyslipidaemia at baseline.

Table 4 shows the adjusted HRs in the sensitivity analyses after modifying some of our assumptions. The reported results did not change when the analyses were restricted to those participants with <6 years of follow-up. Participants in the highest quintile of the DII had a 47 % higher relative risk of developing depression during the first 6 years of follow-up than participants in the lowest quintile. However, when the analyses were restricted to those participants with a depression diagnosis after 2 or 3 years of follow-up, the association was attenuated and it was no longer statistically significant.

Table 4 Sensitivity analysesFootnote * (Hazard ratios (HR) and 95 % confidence intervals)

Ref., referent value.

* The association between extreme quintiles of adherence to the dietary inflammatory index and depression. Adjusted for age, sex, BMI, smoking, physical activity during leisure time, use of vitamin supplements, total energy intake and presence of several diseases at baseline (CVD, diabetes mellitus type 2, hypertension and dyslipidaemia).

† For the five quintiles.

‡ >5648·4 kJ/d (>1350 kcal/d) and <17 405·4 kJ/d (<4160 kcal/d) in men and >4020·8 kJ/d (>961 kcal/d) and <16 970·3 kJ/d (<4056 kcal/d) in women.

§ >3711·2 kJ/d (>887 kcal/d) and <22 781·9 kJ/d (<5445 kcal/d) in men and >5744·6 kJ/d (>1373 kcal/d) and <23 735·8 kJ/d (<5673 kcal/d) in women.

The association between the DII score and depression risk according to categories of several baseline characteristics is shown in Fig. 2. Although none of the interaction terms included in the multivariable models were statistically significant, a stronger association with depression was evident when comparing DII quintile 5 to quintile 1 for participants aged ≥55 years (HR 2·70; 95 % CI 1·22, 5·97) and for those with concomitant diseases including obesity (HR 2·45; 95 % CI 1·15, 5·22); DM2 (HR 4·87; 95 % CI 1·14, 20·81); CVD (HR 2·08; 95 % CI 0·81, 5·36); or a composite of cardiometabolic diseases or risks (HR 1·80; 95 % CI 1·27, 2·57).

Fig. 2 Subgroups analyses. Hazard ratios (HR) (95 % CI)^a for the association between extreme quintiles of adherence to the dietary inflammatory index (DII) and depression. ^aAdjusted for age, sex, BMI, smoking, physical activity during leisure time, use of vitamin supplements, total energy intake and prevalence of several diseases (CVD, type 2 diabetes, hypertension and dyslipidaemia). ^bRepeated measures. Cumulative average for DII (at baseline and after 10 years of follow-up). Energy intake, BMI and prevalence of diseases were also updated. ^cPrevalence of obesity, type 2 diabetes, hypertension or dyslipidaemia.

Discussion

In this analysis from the SUN cohort study, participants with the highest DII score (representing the most pro-inflammatory dietary potential) showed a 47 % higher risk of developing depression compared with participants with the lowest DII (those consuming diets with the greatest anti-inflammatory potential). This result is consistent with that reported in another prospective cohort study, the Nurses’ Health Study, which found a 41 % higher risk for the highest v. the lowest quintile of inflammatory properties of the diet⁽ Reference Lucas, Chocano-Bedoya and Shulze ^{13 )}. Although it used an identical definition of the outcome, the American cohort used a different approach (reduced rank regression) to classify participants according to the inflammatory potential of their dietary patterns.

The results obtained in our analyses confirm those found in other analyses evaluating the role of diet in depression. Numerous prospective studies have reported inverse associations between diet quality or the adherence to prudent or traditional diets and the risk of depression with consistent results across different countries and cultures⁽ Reference Rienks, Dobson and Mishra ^{8 –} Reference Akbaraly, Brunner and Ferrie ^{11 ,} Reference Lai, Hiles and Bisquera ^{31 )}. Similarly, the Mediterranean diet has shown an anti-inflammatory effect⁽ Reference Casas, Sacanella and Urpí-Sardà ^{32 ,} Reference Chrysohoou, Panagiotakos and Pitsavos ^{33 )}. In contrast, Western dietary patterns, characterised by the consumption of processed foods, have been directly associated with depressive disorders⁽ Reference Akbaraly, Brunner and Ferrie ^{11 ,} Reference Le Port, Gueguen and Kesse-Guyot ^{12 )} as well as with elevated levels of some pro-inflammatory markers⁽ Reference Lopez-Garcia, Schulze and Fung ^{34 ,} Reference Fung, Rimm and Spiegelman ^{35 )}.

A large number of studies have reported the possible role of inflammation in depression through mechanisms such as activation of the hypothalamic–pituitary–adrenal axis, tryptophan depletion and decrease in brain-derived neurotrophic factor availability⁽ Reference Zunszain, Hepgul and Pariante ^{36 )}. Cross-sectional evidences from epidemiological studies seem to confirm a bidirectional relationship⁽ Reference Howren, Lamkin and Suls ^{37 )}. Although systemic inflammatory markers have been prospectively associated with depression⁽ Reference Khandaker, Pearson and Zammit ^{38 ,} Reference Pasco, Nicholson and Williams ^{39 )}, not all longitudinal studies have found a significant relationship⁽ Reference Chocano-Bedoya, Mirzaei and O’Reilly ^{40 )}. Thus, the contribution of a pro-inflammatory dietary pattern to the development of depression has not been easy to understand.

One of the most remarkable results obtained in our analysis suggests that the effect of a pro-inflammatory diet (stressor) on depression could be particularly detrimental among individuals with some cardiometabolic conditions (prevalent chronic conditions such as CVD, DM2 or/and obesity) or among those ≥55 years of age. Although the interaction was not statistically significant, its magnitude could be biologically relevant. These results are in accordance with those recently obtained in the PREDIMED trial. In that trial, the adherence to a Mediterranean dietary pattern supplemented with nuts was particularly important to prevent depression among participants with DM2⁽ Reference Sánchez-Villegas, Martínez-González and Estruch ^{41 )}. Similarly, this hypothesis has been suggested in a cross-sectional study conducted recently in Australia, which found that a healthy dietary pattern was associated with a reduced likelihood of depressive symptoms, especially for those with DM2⁽ Reference Dipnall, Pasco and Meyer ^{42 )}.

One possible explanation for our observed results is that the presence of several chronic conditions might lead to maladaptive stress responses within this group, including heightened low-grade inflammation and HPA axis non-habituation. In fact, cytokine levels are strongly affected by socio-demographic and environmental factors such as age, sex, smoking, exercise, obesity or insulin resistance. The link between CVD, inflammation and depressive disorders has been repeatedly suggested. Over the last few years, several studies have established the possible link between inflammation, depression and not only CVD events but also other related conditions such as DM2, MetS or obesity⁽ Reference Hood, Lawrence and Anderson ^{5 ,} Reference Doyle, de Groot and Harris ^{6 ,} Reference Au, Smith and Gariépy ^{43 –} Reference Laake, Stahl and Amiel ^{45 )}. In fact, hyperleptinaemia or insulin resistance in obesity, MetS and DM2 have been linked to inflammatory processes⁽ Reference Flehmig, Scholz and Klöting ^{46 )}, which also are common in depressive disorders. Similarly, obesity (a pro-inflammatory condition⁽ Reference Ouchi, Parker and Lugus ^{47 )}) has been found to be associated with elevated cortisol secretion and higher HPA axis reactivity to psychological stress as well as physiological and pharmacological stimulation⁽ Reference Bjorntorp ^{48 )}. Moreover, in a recent study, McInnis et al.⁽ Reference McInnis, Thoma and Gianferante ^{49 )} found that individuals with higher measures of adiposity showed less efficient HPA axis habituation as well as sensitisation of IL-6 responses to repeated acute stress, indicating that increased adiposity would be related to altered endocrine and IL-6 stress responses. Consistent with our findings, Grosse et al. ⁽ Reference Grosse, Carvalho and Wijkhuijs ^{50 )} reported that monocyte immune activation was not uniformly elevated in all depressive patients, but it was increased only in older subjects. Indeed, the pro-inflammatory effect of the diet could be more relevant among older subjects by inducing sensitisation with increased activation of the inflammatory response system.

Some potential limitations of our study need to be mentioned. Self-reporting of a clinical diagnosis or the use of medication was used as the criteria to establish depression. Our validation study found low sensitivity (0·37) but very high specificity (0·96) for the self-reported diagnosis of depression⁽ Reference Sanchez-Villegas, Schlatter and Ortuno ^{26 )}. Theoretically, with perfect specificity, non-differential sensitivity of disease misclassification will not bias the relative risk estimate⁽ Reference Greenland and Lash ^{51 )}. Similarly, although the validity and reliability of the FFQ have been evaluated⁽ Reference Fernandez-Ballart, Pinol and Zazpe ^{16 ,} Reference De la Fuente-Arrillaga, Vázquez Ruiz and Bes-Rastrollo ^{17 )}, some degree of misclassification may exist in the dietary assessment. However, the use of a cohort design mitigates this to some extent. In this context, misclassification is more likely to be non-differential, and therefore would bias the results towards the null. Another concern is the potential of reverse causation. Participants with subclinical depression at the beginning of our study might have changed their food habits precisely because of their mood disorder. In fact, when the analyses were restricted to those participants who reported a depression diagnosis after 2 or 3 years of follow-up, the association was attenuated. To avoid the possibility that the induction period for the effect of baseline diet might be shorter than the time of follow-up of these ‘late’ cases (some of them diagnosed after 14 years of follow-up), we excluded participants with >6 years of follow-up from the main analyses, updated nutritional data and used cumulative DII after 10 years of follow-up. Both analyses reported similar results, and with an even higher relative risk for depression associated with high DII scores. Another possible weakness is the inability to control for several potential confounders related to psychological features. Finally, our participants are not representative of the general Spanish population. We restricted our cohort to highly educated participants to obtain a better quality of self-reported information, to improve the retention rate and to minimise confounding by educational level, and therefore by socio-economic status⁽ Reference Rothman, Greenland and Lash ^{52 )}.

Several strengths of our study also deserve to be mentioned. They include its large sample size, prospective design, long-term follow-up, the use of updated nutritional data, the ability to control for a variety of major potential confounders, the existence of published validation studies of our assessments and the restriction to highly educated participants who may be able to provide more reliable information.

In conclusion, a higher DII (indicative of a more pro-inflammatory diet) was associated with an increased risk of developing depression among participants from the SUN cohort study. This effect could be even more important among older individuals and those with prevalent comorbidities related to inflammation such as CVD, DM2 or obesity. Further studies analysing the link between inflammation, depression and cardiometabolic conditions are warranted to deepen our understanding about the role of diet in developing depression and other mental disorders.