Association of Poorer Dietary Quality and Higher Dietary Inflammation with Greater Symptom Severity in Depressed Individuals with Appetite Loss

Abstract

Background.

Major depressive disorder (MDD) is the leading cause of years lived with disability; however, little is known about its etiology to inform treatment. For a subset of MDD patients, appetite change and/or bodily inflammation may play a role in exacerbating symptoms. The goal of this study is to examine whether, relative to healthy comparisons (HC), MDD individuals with increased versus decreased appetite symptoms show a differential relationship between diet quality and inflammation.

Methods.

Unmedicated current MDD (n=61) varying in appetite change (decrease (MDD-DE): n=39; increase (MDD-IN): n=22) and HC (n=42) completed 24-hour dietary recall and state depression/anxiety measures. Healthy eating and dietary inflammatory indices were calculated from dietary reports. Blood samples measured five inflammation-related biomarkers. Analyses investigated between- and within-group differences in the Healthy Eating Index (HEI), the Dietary Inflammatory Index (DII), inflammation-related blood biomarkers, and symptom severity.

Results.

While both MDD-DE and MDD-IN exhibited lower HEI scores than HC, only MDD-IN showed higher plasma interleukin-1 receptor antagonist (IL-1RA) and interleukin-6 (IL-6) levels than HC. In contrast, MDD-DE exhibited higher DII scores than MDD-IN and HC. Within MDD-DE, greater symptom severity was associated with lower HEI and higher DII.

Limitations.

Modest sample sizes and the cross-sectional study design limited power to detect within-MDD effects.

Conclusions.

Although MDD, regardless of appetite change, is linked to poorer dietary quality, depression severity was related to dietary characteristics only in subjects who reported appetite loss. Thus, increasing the quality of dietary intake could be a treatment target for some individuals with depression.

Introduction

Major depressive disorder (MDD) is the leading cause of years lived with disability worldwide (Vos et al., 2012) with an estimated economic cost of over $200 billion yearly (Greenberg et al., 2015). About 20-40% of subjects with MDD do not respond to current treatments (Joffe et al., 1996). As various symptom configurations pave the way for an MDD diagnosis, it is probable that biological and environmental contributions to illness course, severity, and treatment response vary substantially as a function of specific symptom profiles (Fried and Nesse, 2015; Kessler et al., 2017). One such symptom marked by substantial heterogeneity between MDD patients is the degree and direction of appetite change during a major depressive episode (MDE), a pattern that appears to be stable within-patients across multiple MDEs (Nierenberg et al., 1996). Recent neuroimaging research demonstrates that neuroendocrine, metabolic, and immune states differentiate two MDD appetite subtypes, wherein increased appetite is linked to heightened blood insulin resistance and inflammation paired with exaggerated brain-bodily signals to food cues, and decreased appetite is characterized by excessive stress hormones paired with attenuated reward responsivity to food images (Simmons et al., 2016; Simmons et al., 2018). In addition, depressed subjects with increased appetite relative to individuals with decreased appetite show greater preference ratings for high-fat and high-sweet food images (Simmons et al., 2016). Patients with atypical depression, which is prominently linked to increased appetite, report lower diet quality than patients with melancholic depression (Rahe et al., 2015). Thus, we hypothesized that MDD subjects with increased appetite would have a higher pro-inflammatory diet and poorer dietary quality intake than MDD patients with decreased appetite. Developing interventions that target these biological dysfunctions may reduce disability for a subset of subjects with MDD, although more research is warranted to determine what proximal factors contribute to these neurochemical alterations.

Diet is one such proximal factor impacting neurochemical changes in the body and a substantial literature shows that optimal nutrition is linked to mental health, including depressive and anxiety disorders (Beydoun and Wang, 2010; Gibson-Smith et al., 2018; Murakami and Sasaki, 2010); in contrast, high intakes of fruit, vegetables, fish, and whole grains are linked to reduced depression risk (Gibson-Smith et al., 2019; Lai et al., 2014). However, more studies need to investigate diet quality in subtypes of clinical MDD patients, including those with increased or decreased appetite subtypes, using assessment tools that conform to federal dietary guidelines. This research is needed for the development of interventions targeting diet changes in federal and state-funded institutions (e.g., schools and other government programs). Utilization of the Heathy Eating Index 2010 (HEI), a comprehensive assessment of diet quality capturing key recommendations of U.S. dietary guidelines (Guenther et al., 2013), may prove beneficial in understanding and developing diet-related treatment targets.

Low-grade inflammation, the chronic production of slight increases in inflammatory markers that alone does not indicate infection (Minihane et al., 2015; Pietzner et al., 2017), is linked to poor diet quality (Ahluwalia et al., 2013; Dias et al., 2015; Fung et al., 2005; van Bussel et al., 2015). While inflammation is thought to play an important role in MDD (Dowlati et al., 2010), the precise cause of the inflammation is unclear (Kiecolt-Glaser et al., 2015). The Dietary Inflammatory Index (DII) is a new literature-derived tool for assessing the inflammatory potential of an individual’s diet (Shivappa et al., 2014). Although recent studies suggest that a greater pro-inflammatory diet (indicated by higher DII scores, consisting of high fat and sugar paired with low fiber, vegetables and fruits) is associated with higher depression symptoms (Phillips et al., 2017; Sanchez-Villegas et al., 2015; Shivappa et al., 2018; Shivappa et al., 2016), several gaps in the literature still need to be addressed to address relationships between nutrition, appetite, inflammation, and MDD. First, prior work assessed dietary intake with the food frequency questionnaire (FFQ), a trait-like measure of food consumption associated with a social desirability bias (Hebert, 2016). Second, most diet-related studies of MDD are solely based on self-report or observational assessments and do not include biological markers of inflammation to determine direct relationships between self-reported eating behavior and physiology (Zimmerman, 1983). Third, extant work includes mixed samples of medicated and unmedicated MDD patients; as some antidepressants may change eating behavior (Fava, 2000), it is unclear whether medication status confounds the relationship between diet and inflammation. Fourth, although self-reported depressive and anxiety symptoms are linked with poor diet quality (Gibson-Smith et al., 2018), the direction of this relationship warrants confirmation by current MDE severity ratings; current symptom severity may be an important third variable paralleling poor diet/appetite and heightened inflammation in MDD.

This study investigated relationships between poor diet, inflammation, and depression by comparing unmedicated MDD subjects, comprised of two appetite subtypes (decrease and increase) to healthy comparison subjects (HC) on dietary nutrition based on federal guidelines (HEI) and current dietary pro-inflammatory states (DII). We employed a well-established dietary measure, the automated self-administered 24-hour recall (ASA24) (Kirkpatrick et al., 2014; Subar et al., 2012), as well as current symptom severity measured by Hamilton Rating Scales for Depression (HAM-D) and Anxiety (HAM-A) (Hamilton, 1959, 1960), and inflammation measured in two ways: (1) five blood-circulating inflammation-related markers; and (2) DII based on 24-hour dietary intake. We hypothesized that MDD patients with increased appetite would endorse poorer dietary quality (lower HEI) paired with greater dietary and biological levels of inflammation (higher DII and blood-based biomarkers) than MDD with decreased appetite and HC. Exploratory analyses investigated whether greater current depression and comorbid anxiety symptom severity (HAM-D and HAM-A) paralleled heightened inflammation and poorer diet quality within subtypes of MDD patients.

Methods

Subjects and Groups

Study recruitment occurred from August 2012 to May 2017 as part of a larger neuroimaging study examining brain circuitry as a function of appetite-related MDD subtypes. Subjects were recruited via radio/internet advertisements and posted fliers. After providing written consent, subjects were screened by trained clinical interviewers to evaluate the following study exclusion criteria: (1) exposure to psychotropic medications within 6 weeks prior to data collection; (2) exposure to medications affecting weight and appetite within 3 months of study participation; (3) use of anti-inflammatory medications (e.g., ibuprofen) for either the study day or the day prior to the study; (4) lifetime eating disorder; (5) serious suicidal ideation or behavior; (6) current psychosis to the extent that the ability to provide informed consent for the study was in doubt; (7) medical conditions or concomitant medications likely to influence cerebral blood flow or neurological function including cardiovascular, respiratory, endocrine and neurological diseases; (8) past year drug or alcohol abuse or lifetime alcohol or drug dependence; (9) current pregnancy; and (10) gastrointestinal disturbances (e.g., inflammatory bowel disease, active peptic ulcer disease, etc.).

Subjects were evaluated in person using an unstructured interview with a psychiatrist as well as the Structured Clinical Interview for DSM-IV-TR (SCID) developed by the American Psychiatric Association (Association, 2000) with trained clinical interviewers. Depression-related changes in appetite and weight are codified as diagnostic markers in the DSM-IV and have long been recognized as cardinal features of depression. Subjects qualifying for the MDD group met DSM-IV-TR criteria for recurrent MDD and were currently in an MDE as verified by the SCID, with HAM-D scores in the moderately-to-severely depressed range (≥18 on the 25-item scale). HC did not meet criteria for any Axis I diagnosis (all diagnostic categories except mental retardation and personality disorder) as verified by the SCID, reported no known first-degree relatives with mood disorders, endorsed a current score on the HAM-D in the “not depressed” range (≤ 7), reported stable weight for the previous 2 months, and were not presently engaged in a weight loss program. Socioeconomic status (SES) was collected using the Hollingshead Four-factor Index of Socioeconomic Status Questionnaire. HC subjects were matched to MDD subjects for body mass index (BMI), age and sex. Once eligible for MDD or HC group assignment, subjects provided written informed consent as approved by the Western IRB and received compensation for their participation.

Data Collection

Subjects arrived at 8am to provide informed consent and consumed a 552-calorie breakfast consisting of an English muffin with eggs and cheese, a side of potatoes and a banana at approximately 8:30am. Next, they completed the ASA24 Dietary Recall (Subar et al., 2012) and were administered the 25-item HAM-D (Hamilton, 1960) and HAM-A (Hamilton, 1959) to index current depression and anxiety symptom severity, respectively. Venous blood was collected in two BD Vacutainer EDTA tubes at noon after 3.5 hours of breakfast completion. Five inflammation-related biomarkers were analyzed from blood plasma: (1) interleukin 1 receptor antagonist (IL-1RA); (2) interleukin 6 (IL-6); (3) interleukin 10 (IL-10); (4) tumor necrosis factor alpha (TNF-α); and (5) C-reactive protein (CRP). Of the 2 EDTA tubes drawn, the tube used for analysis of IL-1RA was centrifuged at 1300g for 10 minutes at room temperature, plasma was removed and aliquoted. A protease inhibitor (Pefabloc SC, Roche Scientific, IN) was added to another tube, which was used for assaying IL-6, IL-10, TNF-α and CRP, and then immediately placed on ice and centrifuged at 1300g for 10 minutes at 4°C prior to removal and aliquoting of plasma. Plasma aliquots were stored at −80°C until analysis. One MDD-IN subject who was on a special diet during the study was excluded. Data from total of 103 subjects (61 MDD, 42 HC) were used for final analyses. Among the 61 unmedicated MDD subjects, 39 reported appetite-decrease (MDD-DE) and 22 reported appetite-increase (MDD-IN) based on their responses to HAM-D appetite questions.¹

Although most of the subjects’ data were included in a previous paper focusing on the appetite changes in depression (Simmons et al., 2018), a higher number of MDD subjects were included in this investigation due to differences in study inclusion criteria. In the previous investigation, appetite change was based on responses to appetite change questions in the mood disorders module of the SCID, which were then confirmed in an interview with a psychiatrist; all depressed subjects included in that study reported either weight gain or loss in their current MDE. For the present study, appetite change was based on HAM-D appetite questions (H4: somatic symptoms gastrointestinal and A3: appetite increase on study day) regardless of subjects’ actual weight change. Specifically, all of the 42 HC subjects, 29 of MDD-DE (74.4%) and 18 of MDD-IN (81.8%) were in the earlier study. Supplementary Table 1 compares the number of subjects assigned to each subgroup obtained using two methods.

HEI and DII Quantification

The total HEI score and its component scores (total fruit, whole fruit, total vegetables, greens and beans, whole grains, dairy, total protein foods, seafood and plant proteins, fatty acids, refined grains, sodium, and empty calories) were calculated according to HEI scoring methods (Guenther et al., 2013) based on dietary intake from the ASA24 Individual Foods and Pyramid Equivalents report (Subar et al., 2012). All the ASA24 dietary recalls were collected for a weekday.

The DII was developed by Shivappa and colleagues (Shivappa et al., 2014) using the following process: 1943 peer-reviewed research articles published through 2010 were scored and weighted based on the effect of diet on six inflammatory markers, including IL-1β, IL-4, IL-6, IL-10, TNF-α and CRP, and the global intake based on consumption data from 11 countries. As a result of this review, a total of 45 pro- or anti-inflammatory food parameters were identified. To calculate the DII, each subject’s dietary intake was: (1) scaled to a z-score based on each food parameter’s global daily mean intake and standard deviation; (2) converted to a percentile score to minimize skew; and (3) centered on 0 and bounded between −1 as maximally anti-inflammatory and +1 as maximally pro-inflammatory to achieve a symmetrical distribution. A DII score for each of the 45 food parameters was calculated from the centered percentile score multiplied by its overall food parameter-specific inflammatory effect score and summed to obtain an overall DII score. Supplementary Table 2 illustrates an example of DII calculations for an individual subject.

Thirty eight food parameters were used in this study: 27 individual nutrient-based food parameters (beta carotene, caffeine, carbohydrates, cholesterol, energy, iron, fiber, folic acid, magnesium, monounsaturated fatty acids, omega-3 fatty acids, omega-6 fatty acids, niacin, protein, polyunsaturated fatty acids, riboflavin, saturated fat, selenium, thiamin, total fat, vitamin A, vitamin C, vitamin D, vitamin E, vitamin B12, vitamin B6, and zinc) and 4 food-item parameters (alcohol, green/black tea, onion, pepper) were obtained from the ASA24 Individual Foods and Pyramid Equivalents and Individual Supplements reports (Subar et al., 2012). Moreover, isoflavones and five flavonoid food parameters (anthocyanins, flavan-3-ol, flavones, flavonols, flavonones) were calculated from the United States Department of Agriculture (USDA) Database for the Isoflavone Content of Selected Foods (Release 2.1, 2015) and Flavonoid Content of Selected Foods (Release 3.1, 2013). Trans fat values were calculated using USDA National Nutrient Database for Standard Reference (Release 28, 2018).²

Inflammation-related Biomarker Immunoassays

Plasma IL-1RA was measured from 98 subjects (59 MDD; 39 HC) with Human IL-1RA Quantikine ELISA kits (R&D Systems, Minneapolis, USA). Plasma samples from 97 subjects (59 MDD, 38 HC) were analyzed for IL-6, IL-10, TNF-α and CRP with V-PLEX Neuroinflammation Panel 1 Human Kit (Meso Scale Diagnostics, Maryland, USA). All plasma samples were tested in duplicate, the intra- and inter-assay coefficients of variation (CV) were 3% and 7% (IL-1RA), 2% and 8% (IL-6), 3% and 10% (IL-10), 2% and 10% (TNF-α), 2% and 11% (CRP), respectively.

Statistical Analyses

All statistical analyses were conducted in R. For demographic characteristics (age, BMI, SES) and clinical (HAM-D, HAM-A) ratings, no outliers were found using z = ±3 across subjects. One-way analysis of variance (ANOVA) tests examined differences between MDD-DE, MDD-IN, and HC. Chi-square tests tested sex differences between groups.

For HEI scores, DII score, plasma IL-1RA, IL-6, IL-10, TNF-α and CRP levels, Shapiro-Wilks tests were employed to test normality of distributions; those that were found to be non-Gaussian were log-transformed. Following Bartlett’s tests for homogeneity of variance, ANOVA was used to assess group differences on IL-6 and TNF-α followed by Tukey’s “Honest Significant Difference” method if the overall test was significant. Even after log-transformation, the distributions for HEI scores, DII score and plasma CRP, IL-1RA and IL-10 were found to be non-Gaussian; therefore, group differences were assessed with Kruskal-Wallis non-parametric tests and followed by Mann-Whitney-Wilcoxon non-parametric post-hoc tests if the overall test was significant. Outliers were defined as z = ±3 across subjects and set as missing. Cohen’s d was computed to evaluate effect size differences between groups.

Pearson’s correlations explored potential relationships between diet (HEI, DII) and symptoms (HAM-D, HAM-A) within the MDD subgroups, and ANOVA tests were used to evaluate slope differences between groups. False Discovery Rate (FDR) correction for multiple comparisons was used for these correlations. Relationships between DII and blood inflammation-related biomarkers within MDD subgroups, HC and all MDD subjects were tested as well.

Results

Demographics

MDD-DE, MDD-IN and HC did not differ on age, BMI, sex or SES, and the two MDD groups did not differ on HAM-D depression severity or HAM-A anxiety severity (Table 1).

Table 1.

Sample Demographics and Clinical Characteristics.

	HC	MDD-IN	MDD-DE	p-value
	Mean (sd)	Mean (sd)	Mean (sd)
N	42	22	39
Age in Years	31.33 (8.56)	31.41 (8.53)	30.13 (9.49)	0.794a
Body Mass Index (kg/m2)	28.37 (5.0)	30.86 (6.42)	29.93 (5.81)	0.205a
Gender = Male (%)	15 (35.7)	5 (22.7)	14 (35.9)	0.512b
Socioeconomic Status Score	39.06 (15.27)	36.00 (17.08)	31.18 (12.96)	0.091a
Hamilton Depression Rating Scale	2.74 (2.56)	27.82 (7.70)	24.38 (6.15)	0.061c
Modified Hamilton Depression Rating Scale	2.43 (2.37)	21.82 (5.64)	21.79 (5.82)	0.988c
Hamilton Anxiety Rating Scale	2.33 (2.91)	17.82 (6.28)	19.33 (7.11)	0.408c

Note. HC, Healthy Control; MDD, Major Depressive Disorder; MDD-IN, Depressed subjects reporting appetite increase; MDD-DE, Depressed subjects reporting appetite decrease; sd, standard deviation; Modified Hamilton Depression Rating Scale: appetite and food related questions were excluded from scoring.

^aOne-way ANOVA tests.

^bχ² test.

^cTwo Sample t-test between MDD-IN and MDD-DE group.

Dietary Quality Results

Both MDD groups endorsed poorer dietary quality, reflected by the lower HEI total score and the empty calories score, than HC (Figure 1A and Figure 1B).

Figure 1.

Healthy eating index (HEI), depression and anxiety symptoms. A. Both major depressive disorder (MDD) subjects with appetite-decrease (MDD-DE) and appetite-increase (MDD-IN) exhibited lower HEI total scores compared to healthy comparison subjects (HC). B. Both MDD-DE and MDD-IN subjects showed lower empty calories scores than HC. C. Within MDD-DE, lower HEI scores were associated with higher depression severity scores (HAM-D). D. Within MDD-DE, lower HEI scores were associated with higher anxiety severity scores (HAM-A).

MDD-DE, MDD-IN and HC groups differed on protein (p = 0.026), total fat (p = 0.030), total fruit (p = 0.040), total vegetables (p = 0.023), whole grains (p = 0.037), seafood and plant proteins (p = 0.001), empty calories (p < 0.001) and HEI total p < 0.001) (Table 2). More specifically, when compared to HC, MDD-DE endorsed lower scores on total fruit (p = 0.031), total vegetables (p = 0.007), whole grains (p = 0.013), seafood and plant protein (p < 0.001), empty calories (p < 0.001) and HEI total (p < 0.001). Similarly, MDD-IN reported lower scores than HC on total fruit (p = 0.037), empty calories (p = 0.002) and HEI total (p = 0.009). Finally, with respect to differences between MDD groups, MDD-DE showed lower protein (p = 0.012) and total fat intake (p = 0.016) than MDD-IN. On the whole, Cohen’s d effect sizes for significant group differences ranged from medium to large (Table 2).

Table 2.

Mean Food Intake and Healthy Eating Index (HEI) Scores.

	HC (n=42)	MDD-IN (n=22)	MDD-DE (n=39)	p-value	HC vs. MDD-IN		HC vs. MDD-DE		MDD-IN vs. MDD-DE
	Mean (sd)	Mean (sd)	Mean (sd)		p-value	db	p-value	db	p-value	db
Food Intake
Energy(kcal)a	2336.42 (1111.66)	2756.56 (1746.95)	1995.89 (1136.02)	0.110
Protein(g)a	92.88 (49.1)	97.53(41.30)	74.72(59.14)	0.026	0.419	0.10	0.046	0.34	0.012	0.43
Total Fat(g)a	95.56 (46.80)	114.85 (68.71)	77.11 (43.41)	0.030	0.229	0.35	0.066	0.41	0.016	0.70
Carbohydrates(g)a	267.45 (121.28)	338.13 (259.16)	246.97 (146.46)	0.317
HEI Subscores
Total Fruita	2.22 (2.12)	0.99 (1.33)	1.29 (1.86)	0.040	0.037	0.65	0.031	0.46	0.884	0.18
Whole Fruita	2.01 (2.21)	1.27 (1.92)	1.05 (1.83)	0.081
Total Vegetablesa	3.27 (1.65)	2.65 (1.49)	2.27 (1.53)	0.023	0.142	0.38	0.007	0.63	0.359	0.26
Greens and Beansa	1.83 (2.27)	1.05 (1.85)	1.12 (1.84)	0.119
Whole Grainsa	2.58 (2.74)	1.82 (2.79)	1.54 (2.49)	0.037	0.217	0.27	0.013	0.40	0.239	0.11
Dairya	5.96 (2.97)	5.20 (3.43)	4.70 (3.39)	0.203
Total Protein Foodsa	4.61 (1.08)	4.50 (1.24)	4.20 (1.66)	0.536
Seafood/Plant Proteins a	2.04 (2.18)	1.47 (2.10)	0.67 (1.34)	0.001	0.241	0.26	<0.001	0.75	0.041	0.49
Fatty Acidsa	4.50 (3.44)	3.62 (3.43)	3.38 (3.45)	0.173
Refined Grains a	6.16 (3.41)	6.60 (2.95)	6.35 (3.68)	0.882
Sodiuma	3.94 (3.67)	3.86 (3.42)	5.44 (3.49)	0.108
Empty Calories a	11.11 (6.39)	6.04 (5.18)	5.11 (5.690)	<0.001	0.002	0.84	<0.001	0.99	0.292	0.17
HEI Total Scorea	50.22 (16.29)	39.08 (8.59)	37.10 (10.82)	<0.001	0.009	0.79	<0.001	0.94	0.351	0.20

Note. HC, Healthy Control; MDD, Major Depressive Disorder; MDD-IN, Depressed subjects reporting appetite increase; MDD-DE, Depressed subjects reporting appetite decrease.

^ap-values are from nonparametric results using the Kruskal-Wallis test (Wilcoxon test for the two-group case).

^bCohen’s d effect sizes computed with the effsize package in R.

Statistically significant p-values are indicated in bold font.

Lower HEI scores were associated with higher HAM-D scores (r = −0.38, p_corrected = 0.024) in MDD-DE subjects but no such relationship was observed in MDD-IN subjects (Figure 1C). In addition, MDD-DE and MDD-IN group differences were observed in the slope of the relationship between HEI and HAM-D (F(_{1, 57}) = 4.48,p) = 0.039). Fisher’s r-to-z transformations were applied to this correlation for each MDD subtype and then compared; results indicated that the relationship between HEI and HAM-D was significantly more negative in MDD-DE than MDD-IN, z = −2.16, p = 0.031.

Similarly, there was a negative association between HEI and HAM-A (r = −0.38, p_corrected = 0.024) in the MDD-DE group but not in the MDD-IN group (Figure 1D). In addition, there was a trending group difference in the slope of relationship between HEI and HAM-A (F(_{1, 57}) = 3.78,p) = 0.056). Fisher’s r-to-z transformations were applied to this correlation for each MDD subtype and then compared; results indicated that the relationship between HEI and HAM-A was significantly more negative in MDD-DE than MDD-IN, z = −2.05, p = 0.040.

Dietary and Blood-Based Inflammation Results

We observed group differences between MDD-DE, MDD-IN and HC on DII (p = 0.001) and blood inflammation-related biomarkers IL-1RA (p 0.048) and IL-6 (p = 0.004), but there were no group differences in CRP, TNF-α and IL-10 (Table 3). MDD-IN endorsed higher levels of plasma IL-1RA (p = 0.029) and IL-6 (p = 0.005) than HC (Figures 2A and 2B). In comparison, MDD-DE showed higher DII scores than HC (p = 0.001) and MDD-IN (p = 0.025) (Figure 2C).

Table 3.

Blood and Dietary Inflammation.

	HC (n=42)	MDD-IN (n=22)	MDD-DE (n=39)	p-value	HC vs. MDD-IN		HC vs. MDD-DE		MDD-IN vs. MDD-DE
	Mean (sd)	Mean (sd)	Mean (sd)		p-value	d	p-value	d	p-value	d
CRP (mg/L)b	2.15 (1.94)	5.35 (5.66)	3.85 (5.41)	0.073
IL-1RA (pg/mL)b	351.36 (281.38)	537.60 (434.95)	340.64 (434.95)	0.048	0.029	0.54	0.174	0.04	0.091	0.64
aIL-6 (pg/mL)c	−0.11 (0.62)	0.44 (0.69)	0.24 (0.63)	0.004	0.005	0.85	0.051	0.56	0.488	0.30
aTNF-α (pg/mL)c	0.75 (0.27)	0.80 (0.27)	0.79 (0.38)
IL-10 (pg/mL)b	0.31 (0.15)	0.54 (0.43)	0.53 (0.82)	0.061
DIIb	2.80 (2.68)	3.03 (2.05)	4.74 (2.29)	0.001	0.935	0.09	0.001	0.77	0.025	0.77

Note. HC, Healthy Control; MDD, Major Depressive Disorder; MDD-IN, Depressed subjects reporting appetite increase; MDD-DE, Depressed subjects reporting appetite decrease. CRP, C-reactive protein; IL-1RA, Interleukin 1 receptor antagonist; IL-6, Interleukin 6; TNF-α, Tumor necrosis factor alpha; IL-10, Interleukin 10. DII, dietary inflammatory index. Plasma IL-6, TNF-α, CRP and IL-10 from 6 subjects (4 HC, 2 MDD-DE) and IL-1RA from 5 subjects (3 HC, 2 MDD-DE) were not measured due to failure of blood collection, missing blood samples or date of the collection; One subject from HC, one subject from MDD-DE and one subject from MDD-IN were excluded for CRP due to their z-score > 3; Two MDD-DE subjects were excluded for IL-1RA due to their z-score > 3; One MDD-DE subject was excluded for IL-10 due to its z-score > 3; One MDD-DE subject were excluded for IL-10 due to its z-score > 3.

^aLog-transformed, natural base e was used for log transformation.

^bp-values are from nonparametric results using Kruskal test (Wilcox test for two-group case).

^cOne-way ANOVA followed by Tukey’s ‘Honest Significant Difference’ method.

Cohen’s d effect sizes computed with the effsize package in R.

Statistically significant p-values indicated in bold font.

Figure 2.

Plasma inflammation-related markers, dietary inflammatory index (DII) and anxiety severity. A. Major depressive disorder (MDD) subjects with appetite-increase (MDD-IN) exhibited higher Interleukin 1 receptor antagonist (IL-1RA) levels than HC. B. MDD-IN subjects exhibited higher log-transformed plasma Interleukin-6 (IL-6) levels than HC. C. MDD-DE subjects showed higher dietary inflammatory index (DII) scores compared to HC and MDD-IN. D. Higher DII scores were associated with higher anxiety severity scores in MDD-DE, while higher DII scores were associated with lower anxiety severity scores in MDD-IN.

In MDD-DE subjects, higher DII scores were associated with more severe HAM-A ratings (r = 0.45, p_corrected = 0.016, Figure 2D), and a group difference was evident in the slope of the relationship between DII and HAM-A (F(_1,57) = 13.28, p < 0.001). In the MDD-IN group, those individuals who had a higher DII were rated as having a lower HAM-A score (r = −0.48, p = 0.023). Fisher’s r-to-z transformations were applied to this correlation for each MDD subtype and then compared; results indicated that the relationship between DII and HAM-A was significantly more positive in MDD-DE than MDD-IN, z = 3.55, p < .001.

Supplementary Figure 1 depicts correlations between HEI, DII, and blood-based biomarkers for each group separately. Among the five-blood inflammation-related biomarkers, lower DII was associated with lower IL-1RA within HC (r = 0.35, p = 0.028) but no correlations between DII and IL-1RA were significant for the MDD groups (Supplementary Figure 1).

Discussion

The goal of this study was to determine whether unmedicated MDD subjects who report an increase in appetite during their major depressive episode relative to those who report a decrease in appetite and healthy comparison subjects show a differential relationship between dietary quality and inflammation. Contrary to our hypothesis that MDD-IN would exhibit poorer diet and greater inflammation than MDD-DE and HC, we did not find a differential relationship. Instead, there were three main findings. First, both MDD-IN and MDD-DE endorsed poorer dietary quality than HC, consisting of more empty calories (solid fats, alcohol, and added sugars) as well as lower consumption of total fruits. Second, the MDD-IN group exhibited greater blood plasma markers of inflammation, IL1RA and IL-6, than HC, while elevated dietary inflammation was only observed in the MDD-DE group. Moreover, dietary inflammation was linked to IL-1RA in HC subjects, but it was unrelated to blood markers of inflammation in the MDD subgroups or MDD as a whole. Third, poorer dietary quality was associated with greater current depression and anxiety severity only within the MDD-DE group; similarly, current anxiety severity was linked to higher dietary inflammation only in the MDD-DE group. Taken together, these results support the idea that dietary quality may be a modifiable treatment target for individuals who report a decrease in appetite during their major depressive episode.

Based on earlier studies (Paans et al., 2018; Rahe et al., 2015; Simmons et al., 2016) indicating that atypical depression and depression with increased appetite are associated with an unhealthy eating style, we hypothesized that MDD-IN would exhibit worse overall dietary quality than MDD-DE and HC; however, this hypothesis was inaccurate, as both MDD groups endorsed greater consumption of non-nutritious food than HC. Although dietary quality was not related to symptoms in MDD with increased-appetite, within MDD with appetite loss, patients who felt the worst during their current MDE also possessed the least healthy diets. A recent review article shows that higher dietary quality is associated with lower risk for the onset of depressive symptoms, but not all available results are consistent with the prediction that diet influences depression risk (Molendijk et al., 2018). For instance, a recent large randomized clinical trial from the European MoodFood project indicates that nutritional strategies (multi-nutrient supplementation or food-related behavioral activation therapy aimed at increasing diet quality) does not prevent MDD onset (Bot et al., 2019). Interestingly, in our study population, if we take MDD subjects as a whole regardless of appetite change, no relationships were found between diet quality and symptom severity (Supplementary Figure 1C). Although prior work suggests that appetite change (increase or decrease) moderates the relationship between depression severity and quality of diet (Gibson-Smith et al., 2016; Martin et al., 2011), our study does not support the association between poorer dietary quality and worse mood/anxiety symptom severity across appetite subtypes of MDD, but only in MDD subjects with current appetite loss. Taken together, although appetite-related MDD subtypes report diets that are similarly poor, the novel findings presented here strongly suggest that improving dietary quality may lower overall symptom severity in those for whom depression presents with appetite loss. Improving dietary quality includes increasing intakes of fruit, vegetables, whole grains, seafood and plant proteins paired with lowering intakes of refined grains, sodium and added sugar (USDA, 2015).

Previous studies demonstrate that a pro-inflammatory diet is associated with greater risk of physical and mental health problems, including heightened depressive symptoms and/or disrupted executive functions (Akbaraly et al., 2016; Esteban-Cornejo et al., 2018; Frith et al., 2018; Phillips et al., 2017; Sanchez-Villegas et al., 2015; Shivappa et al., 2018; Shivappa et al., 2016); a recent review article also indicates that diet explains elevated levels of inflammatory markers observed in severe mental illness (Firth et al., 2019). In our study, in addition to dietary inflammation based on self-report, we tested five inflammation-related biomarkers, including IL-1RA, IL-6, TNF-α, CRP and IL-10. MDD subjects with increased-appetite possessed significantly higher IL-1RA and IL-6 plasma levels than HC; in contrast, TNF-α, CRP and IL-10 did not differ between groups. Heightened IL1RA and IL-6 findings in our MDD within our increased-appetite group are consistent with previous work (Lamers et al., 2013; Simmons et al., 2018); however, similar levels of dietary inflammation between MDD with increased-appetite and healthy controls indicate that the elevated blood inflammation in depression with increased-appetite is not explained by dietary inflammation.

Previous studies investigating the association between DII and blood inflammation-related biomarkers in adults and adolescents show that diet as a whole plays an important role in modifying inflammation (Shivappa et al., 2017; Shivappa et al., 2015). A positive relationship between DII and IL-1RA was observed in our healthy subjects but not in the MDD subgroups. As depression symptoms such as loss of concentration may lower the performance of ASA24 dietary recall, therefore affecting the DII score, a more direct measure of diet and inflammation is needed within MDD subjects to re-evaluate links between self-report and blood-based indicators of inflammation.

We believe that this study is the first to report differences in dietary inflammation between MDD subtypes with different appetite changes. Although studies suggest that higher dietary inflammation is associated with worse depression symptoms (Phillips et al., 2017; Sanchez-Villegas et al., 2015; Shivappa et al., 2018; Shivappa et al., 2016), one additional study finds no association between inflammatory dietary patterns and depressive symptoms in Italian older adults (Vermeulen et al., 2018). One possibility is that the link between dietary inflammation and depression is driven by the decreased-appetite group; in our study sample, 53% of the unmedicated MDD subjects tend to lose their appetite, while only 30% of MDD tend to eat more. Given that our results demonstrated a link between higher dietary inflammation and higher anxiety severity within MDD with decreased-appetite, reducing dietary inflammation in this subset of MDD may help to reduce anxiety symptoms. Reducing dietary inflammation includes lowering intake of vitamin B12, carbohydrate, cholesterol, energy, total fat, saturated fat, and trans-fat, paired with increasing intake of vitamin B₁₂ carotene, eugenol, fiber, folic acid, garlic, ginger, niacin, omega-3 and omega-6 fatty acids, onion, pepper, turmeric, vitamins A, C, D and E, green/black tea, flavonoids and isoflavones (Shivappa et al., 2014). Multiple dietary changes as outlined above may be associated with even larger decreases in depressive symptom severity and/or MDEs, although much more research is warranted to test this hypothesis. Additional longitudinal studies testing dietary interventions in depressed patients with decreased-appetite are needed to determine whether changes in food/nutrient quality improve clinical outcomes and quality of life.

This study has several strengths, including a sample of unmedicated MDD patients, variability in appetite change with similar BMI and SES across groups, two metrics of inflammation, and dietary index measures that correspond to FDA guidelines. However, this investigation has limitations that warrant discussion. First, individual differences in the degree of regular physical exercise and percentage of body fat may have influenced blood inflammatory biomarkers; unfortunately, we did not collect fitness data or percentage body fat on all subjects to evaluate this hypothesis. Second, the cross-sectional nature of this investigation limits conclusions that can be drawn regarding directionality of relationships between diet, inflammation, and psychopathology. Longitudinal studies are needed to clarify direction of causality.

Conclusions

Despite these limitations, this investigation demonstrates that: (1) MDD patients currently experiencing a major depressive episode with decreased-appetite have unhealthy, pro-inflammatory diets that scale with severity of their depression/anxiety symptoms; and (2) MDD patients with increased-appetite have elevated bodily inflammation that is not associated with pro-inflammatory diets. Further studies are needed to evaluate whether dietary changes produce reductions in depressive symptoms for at least a subset of patients experiencing poor dietary quality and high dietary inflammation.

Highlights:

• Depression is linked to poor dietary quality independent of appetite change.

• Depression with increased appetite relates to heightened plasma inflammation.

• Depression with decreased appetite relates to heightened dietary inflammation.

• Depression and anxiety severity correlate with dietary quality in those with appetite loss.

• Anxiety severity correlates with dietary inflammation in those with appetite loss.