Adherence to a Vegetarian Diet is not Associated With Depression: Results From the National Health and Nutrition Examination Surveys

Article information

Psychiatry Investig. 2023;20(4):315-324

Publication date (electronic) : 2023 April 20

doi : https://doi.org/10.30773/pi.2022.0251

Maximilian Andreas Storz^,1, Alvaro Luis Ronco²

¹Department of Internal Medicine II, Center for Complementary Medicine, Freiburg University Hospital, Faculty of Medicine, University of Freiburg, Freiburg, Germany

²Unit of Oncology and Radiotherapy, Pereira Rossell Women’s Hospital, Montevideo, Uruguay

Correspondence: Maximilian Andreas Storz, MD Department of Internal Medicine II, Center for Complementary Medicine, Freiburg University Hospital, 55, Hugstetter Str., Freiburg 79106, Germany Tel: +49 157 545 43852, Fax: +49 761 270 83230 E-mail: maximilian.storz@uniklinik-freiburg.de

Received 2022 September 2; Revised 2022 October 17; Accepted 2023 January 11.

Abstract

INTRODUCTION

Over the past decades, the interest in vegetarian and vegan dietary patterns has increased in the general population and within the scientific community [1,2]. Ethical, environmental, and health motivations are the main drivers of this development [3-6]. Adherence to vegetarian and vegan diets has also been associated with numerous physical health benefits, including a lower risk for cardiovascular disease [7], improved glycemic control in type-2-diabetes, and a more favorable body weight [8,9]. Notably, vegetarian mental health is less well documented [10], and the number of studies investigating potential associations between vegetarian diets and depression is limited [11,12].

Some studies suggested that a vegetarian diet is associated with a lower risk of depression and does not adversely affect mood states, whereas other studies reported opposite findings, highlighting differences in the prevalence of depressive symptoms between vegetarians and omnivores [10,13-15].

In nutritional psychiatry, studies suggested positive associations between a higher intake of fruits, vegetables, and whole grains and a lower depression risk [13,16,17]. Diets high in sugar and processed fatty foods, on the other hand, have been associated with an increased risk for depression [18-20]. A 2020 analysis showed a significant association between red and processed meat intake and depression risk, reporting an overall effect size of 1.08 (95% confidence interval [CI]=1.04–1.12, p<0.001) [21]. In contrast, a 2021 review-supported by the National Cattlemen’s Beef Association-suggested that vegetarians are more likely to suffer from major depression [22]. Regrettably, the authors did not report an overall effect size of depression for red and processed meat. Of note, a recently published meta-analysis by Askari et al. [23] could not confirm the aforementioned association. The authors found no significant associations between the consumption of a vegetarian diet and depression.

Whether vegetarian diets are associated with depression is thus subject to an ongoing and controversial debate. Additional studies investigating potential association between vegetarian diets and depression are therefore warranted [24]. Using data from the National Health and Nutrition Examination Surveys (NHANES), the present analysis investigated whether adherence to a self-reported vegetarian diet was associated with depression in a nationally representative sample of US adults.

METHODS

Outcome variable and covariables

For the present analysis, we merged demographic data, examination data and questionnaire data. The respective NHANES modules and their usage are explained hereafter.

We assessed depression (outcome variable) using the Patient Health Questionnaire (PHQ-9). The PHQ-9 is a nineitem screening instrument that covers questions about the frequency of symptoms of depression over the past 2 weeks [33]. Response categories to each question included “not at all,” “several days,” “more than half the days,” and “nearly every day” and were given a score ranging from 0 to 3. Thus, potential total scores range from 0 to 27. The PHQ-9 is a frequently used tool in the primary care setting and considered reliable in the diagnosis of depression in the general population [29,33-35]. We treated the data dichotomously and defined depression as a PHQ-9 score of 10 or higher [29]. This cut-off point has been validated and deemed suitable in previous studies [36,37]. A-cut off score of 10 or above maximizes combined sensitivity and specificity [38]. We defined individuals with a score ≤9 as having no depression.

Our covariates included sociodemographic factors known to be associated with depression and/or impaired mental health (e.g. female gender, smoking history, elevated C-reactive protein [CRP] levels). For this, we carefully screened the literature and checked previously published publications that used comparable datasets [29].

Demographic data included: sex, age, race/ethnicity, marital status, annual household income, and education level. Race and ethnicity included the following categories: Mexican American, Other Hispanic, Non-Hispanic White, Non-Hispanic Black, and Other Race [39]. Marital status included the following categories: married, widowed, divorced, separated, never married, and living with a partner. Annual household income comprised two categories: under $20,000 and over $20,000. NHANES education level categories were not modified and can be obtained from Table 1. Body mass index (BMI) was obtained from the body measures dataset and categorized into four categories: obesity (BMI ≥30), overweight (BMI 25.00–29.99), normal weight (BMI 18.50–24.99), and underweight (BMI ≤18.49).

Table 1.

Demographic, anthropometric, behavioral, and clinical characteristics by depression status in individuals aged 20 years or older (weighted)

CRP levels, a classical acute-phase reactant and one of the best measures of the acute-phase response to an infectious disease or other causes of tissue inflammation, were obtained from the laboratory dataset. CRP was quantified by latex-enhanced nephelometry and categorized in two groups (based on an approach by O’Neil et al.29): below 0.5 mg/dL or ≥0.5 mg/dL.

Vegetarian status was obtained from the NHANES diet behavior and nutrition section, which provides personal interview data on various dietary behavior and nutrition-related topics. This component included the question “Do you consider yourself to be a vegetarian?” that we used for the group assignment. Assessment of vegetarian status was thus self-reported and based on a yes/no question [40].

Smoking status was assessed using data from the NHANES smoking (cigarette use) module and categorized based on a method by O’Neil et al. [29] We differentiated “current smokers” from “former smokers” and “never smokers,” based on the questions “Do you now smoke cigarettes?” and “Have you smoked at least 100 cigarettes in your entire life?”

Statistical analysis

We used Stata 14 statistical software (StataCorp. 2015. Stata Statistical Software: Release 14; StataCorp LP, College Station, TX, USA) for the entire statistical analysis. We used Stata’s “. svyset” and “. svy” commands to account for the complex NHANES survey design characteristics and the population weights. Prior to the analysis, we generated an appropriate 4-year-sample-weight for the Mobile Examination Center data. Unconditional subclass analyses (preserving the main survey design and providing larger standard errors) were performed to compare demographic, anthropometric, and examination data between depressed and non-depressed individuals (as assessed by the PHQ-9 questionnaire) [41].

To build an adequate logistic regression model, we followed the model-building steps proposed by Heeringa et al. [41] and Hosmer et al. [42] In a first step, we conducted exploratory bivariate analyses to check the eligibility of potential candidate predictors (other than vegetarian status) [41]. We investigated the association between PHQ-9-defined depression and vegetarian status (and other demographic and anthropometric predictors, respectively) using Rao-Scott F-tests of associations (Table 1). Candidate predictors of scientific interest and a bivariate relationship of significance p<0.25 with the response variable were included in the multivariate logistic model (Table 2). All variables were entered as categorical variables into the regression models.

Table 2.

Multivariate logistic regression models examining potential associations between selected predictor variables and the odds for depression

In a first model, we investigated whether PHQ-9-defined depression was associated with vegetarian status after adjusting for the effects of various demographic factors (sex, age, race/ethnicity, annual household income, and marital status). Smoking status, educational level, BMI, and CRP were then added in a second model to examine whether the relationship was independent of these.

The aforementioned covariables were identified based on the literature and adjustment was performed because individuals on a vegetarian diet oftentimes exhibit a healthier lifestyle [43], characterized by lower smoking rates [44] and lower biomarkers of systemic inflammation [45]. Odds ratios (OR) and their CIs are presented in Table 2. We considered a p-value of less than 0.05 statistically significant. All tests were twotailed. Ultimately, we used Archer and Lemeshow’s [46] designadjusted test to assess the Goodness-of-fit for our logistic regression models.

RESULTS

Our analysis included 9,584 individuals with a complete dataset, of which 212 were self-reported vegetarians (unweighted). A total of 910 individuals (9.49%) had PHQ-9 scores ≥10, suggesting depressive symptoms. A total of 8,674 individuals (91.51%) reported PHQ-9 scores <10, suggesting no depressive symptoms.

More than 65% of individuals with PHQ-9 scores ≥10 were female and more than 27% were between 45 and 54 years of age. Almost 64% were non-Hispanic White and 40.25% were married. Approximately 70% had an annual household income over $20,000. Only 1.24% of individuals suffering from PHQ-9-defined depression were self-reported vegetarians. A large proportion of individuals with PHQ-9 scores ≥10 had an elevated body weight: more than one-forth were overweight and almost 44.60% were obese. Table 1 shows the other demographic, anthropometric and clinical data characterizing our sample. The weighted proportions describing our sample characteristics may also be obtained from Table 1.

Self-reported vegetarians differed from non-vegetarians with regard to the male to female ratio (32.78% vs. 67.22%, vs 49.35% vs. 50.65%) and with regard to the proportion of individuals with a pathologically elevated body weight (BMI >25, 52.10% vs. 70.09%). The proportion of current smokers was significantly smaller among self-reported vegetarians (11.03% vs. 21.59%), whereas the proportion of participants with a college degree or above was significantly higher (43.31% vs. 27.12%). No significant differences between self-reported vegetarians and non-vegetarians were found with regard to mean age, annual household income, and marital status.

After adjustment for sex, age, race/ethnicity, annual household income, and marital status, vegetarian status was not associated with increased odds for PHQ-9-defined depression (Table 2). To the contrary, we found lower odds (OR: 0.49 [CI: 0.24–0.98], p=0.047) for PHQ-9-defined depression in individuals reporting a vegetarian diet.

After additionally adjusting for educational level, BMI, CRPlevels, and smoking status (Table 2), vegetarian status was no longer associated with PHQ-9-defined depression. Other factors that increased the odds for PHQ-9-defined depression include female sex, Hispanic race as well as divorce and separation status. From the covariates in the second model, obesity, smoking and CRP increased the odds for PHQ-9-defined depression in the employed model.

Archer and Lemeshow’s [46] design-adjusted test to assess the goodness of fit for both of our models revealed p-values of 0.26 (FA-L[9,24]=1.35) and 0.99 (FA-L[9,24]=0.99), respectively–suggesting that the null hypothesis stating that the models fit the data well is not rejected.

DISCUSSION

The present study sought to investigate whether self-reported vegetarianism was associated with PHQ-9-defined depression in a nationally representative sample of the US civilian non-institutionalized population. After adjustment for various sociodemographic factors in multivariate logistic regression, we did not find an association between both entities. A discussion in the context of previous studies and potential pathomechanisms as well as explanations for our findings is thus warranted.

Several studies suggested a potential relationship between dietary patterns and mental health [13]. Dietary approaches limiting the intake of processed foods and emphasizing whole grains, vegetables, and fruits were associated with a reduced likelihood of depression [47,48].

A recently published prospective cohort study by Shen et al. [13] demonstrated that a Taiwanese vegetarian diet may reduce the odds for depression. The authors followed more than 12,000 participants from the Taiwanese Buddhist Tzu Chi Foundation since 2005. The vegetarian group in this particular cohort had a lower incidence of depressive disorders (2.37 vs. 3.21 per 10,000 person-years) than non-vegetarians. A landmark study in the field of nutritional psychiatry, this was the first prospective study investigating the long-term effects of vegetarian diet on the development of depressive disorders. Main strengths included the large sample size, the robust study design, and the assessment of diet through a reliable and validated 57-item food frequency questionnaire [13,49].

A previously published Austrian matched sample study revealed opposite findings [15]. Burkert et al. [15] found significantly higher rates of mental illness (including anxiety disorders or depression) in vegetarians (9.4% vs. 5.8% in individuals who consumed carnivorous diets rich in meats). The results, however, were also based on cross-sectional data, and the authors acknowledged that no statements were possible as to whether poorer mental health in vegetarians was the result of the chosen diet or whether the diet was consumed due to poorer health.

Li et al. [50] also reported that a vegetarian diet may be associated with depressive symptoms in older Chinese men. Compared to men consuming a meat-based diet, those who consumed a vegetable-based diet had a higher rate of depression (OR [CI]: 1.62 [1.07–2.46]), and severer depression symptoms. Yet again, the cross-sectional design of the study did not allow for determining causality. Likewise, Hibbeln et al. [12] reported a higher frequency of depressive symptoms in vegetarian men after adjustment for sociodemographic factors. Vegetarians in this particular cohort from South West England had higher average depression scores than non-vegetarians (mean difference: 0.96 points [CI: +0.53 to +1.40]). The authors discussed nutritional deficiencies as a potential cause but also acknowledged that reverse causation could not be ruled out.

More recently, Jin et al. [51] investigated the association between a vegetarian diet and depression among South Asians in the US. The authors reported a 43% lower odds of depression among vegetarians (p=0.023) and called for longitudinal examinations confirming their findings.

Unfortunately, the number of longitudinal intervention studies in this particular field is limited. One example for a sound intervention study in this field is the trial by Kahleova et al. [52] The authors reported significant improvements in depressive symptoms (using Beck Depression Inventory) after a 24-week vegetarian intervention in 74 individuals suffering from type-2-diabetes. A 2012 study by Beezhold and Johnston [53] investigated mood changes following a 2-week dietary intervention. The authors randomized omnivorous individuals to either a lacto-ovo-vegetarian group (avoiding meat, fish, and poultry) or to a group consuming fish 3–4 times weekly (avoiding poultry and meat) or to a control group consuming meat, fish, and poultry daily. Those who adopted the lacto-ovo-vegetarian diet significantly improved their mood, whereas the other two groups did not.

There are many potential explanations how a vegetarian diet could affect mood and depression risk. It is now widely accepted that depression is associated with chronic and systemic low-grade inflammation [54,55]. Individuals suffering from depression exhibit significantly higher concentrations of proinflammatory cytokines, including tumor necrosis factor-α and interleukin (IL)-6 [56]. A recently published meta-analysis confirmed depression as a pro-inflammatory state and emphasized the role of elevated inflammatory markers in depression (particularly CRP and IL-12) [57]. Vegetarian and vegan diets are abundant in antioxidants, polyphenols and low in pro-inflammatory precursors, such as arachidonic acid [17]. They have been associated with favorable reductions in inflammatory biomarkers in various studies [58,59], and appear to protect against inflammation and hence, potentially depression [13].

According to Shen et al. [13], vegetarian diets convey mood protection via a reduction of oxidative stress. Oxidative stress is the consequence of an imbalance between the production of reactive oxygen species and their removal. A moderate production and activity of oxygen species is necessary for body health [60]. Pathologically elevated levels, however, are detrimental and play an important role in depressive disorders [61]. A regular consumption of antioxidant-rich plant-foods has been recommended to increase the body’s antioxidative [17,62]. A vegetarian diet contains more antioxidant vitamins (betacarotene, vitamin C, and vitamin E) than an omnivorous diet and has been associated with lower oxidative stress [63,64]. This could be one explanation why some studies reported lower odds of depression in vegetarians [13,51].

The reservation must be made, however, that not all vegetarian diets are created equal. The Taiwanese vegetarian diet, as reported in the Tzu Chi Vegetarian Study and Tzu Chi Health Study, excluded meat and fish and included only few dairy products [65]. Moreover, the daily amount of animal protein intake in those cohorts was quite low and Taiwanese vegetarians differ from Western vegetarians in many ways [66]. The vegetarian diet consumed by vegetarians in our sample was different. Several participants occasionally reported consumption of some type of flesh products (meat, poultry, and/or seafood), and the amount of eggs and dairy was higher in this cohort as compared to the Taiwanese studies [67]. This must be kept in mind when comparing our investigation with the one by Shen et al. [13] and could partly explain the lack of an association between vegetarian diet status and PHQ-9-defined depression in our sample.

It is often difficult to disentangle whether the reported poorer mental health in vegetarians in some studies was the result of the chosen diet or whether the diet was consumed due to poorer mental health. However, one must keep in mind that an unbalanced and restrictive vegetarian diet may result in nutritional deficiencies. [15]. Said potential deficiencies, particularly in omega-3-fatty acids, vitamin B 12, iron, and folate, could be a potential explanation for the higher odds of depression in vegetarians observed in several studies [12,15,50].

Finally, it is of paramount importance to consider lifestyle factors and behaviors when discussing about vegetarian diets and their association with depression. Vegetarians frequently yield a healthier lifestyle with higher levels of physical activity as compared to non-vegetarians [68]. Some studies also suggest that vegetarians are also less likely to smoke [69], and are more likely to have a higher educational level as compared to meateaters [70]. Adjustment for these co-variables is warranted in logistic regression models and frequently done [51]; yet not all studies considered a complete set of factors potentially impacting depression [15]. This could partially explain the aforementioned contradictions between studies and should be carefully considered in future investigations.

In our case, the initially significantly reduced OR in model 1 was not retained when adjusting for lifestyle factors in model 2. Said phenomenon could be interpreted in multiple ways. One potential explanation is that BMI and inflammation (as assessed by CRP levels) could be potentially mediators instead of confounders. As discussed earlier, vegetarian diets may decrease systemic inflammation and associate with a healthier body weight. It is not inconceivable that when we adjusted for said (hypothetical) mediators, we were potentially blocking “the causal” path from diet to depression, resulting in the change in significant OR.

Article information Continued