Meals that heal: a randomized controlled trial testing the feasibility of commercial meal delivery as a convenient dietary intervention for depression

Abstract

Minimally processed dietary interventions have been associated with reductions in depressive symptoms; however, the effort required to plan, shop for, and prepare minimally processed foods may limit adherence. Providing convenient access to minimally processed meals through commercial delivery services may reduce these barriers and potentially enhance intervention impact. This pilot randomized study evaluated the feasibility, acceptability, and preliminary effectiveness of this novel approach through a two-week dietary intervention among adults with moderate to moderately severe depressive symptoms. After one week of baseline observation, participants (N = 31) received nutritional guidance on how to eat a more minimally processed diet for two weeks, which was either fully facilitated by commercial meal delivery (n = 20), or self-implemented (e.g., nutritional guidance; n = 11). Dietary quality (assessed via Diet ID) and depressive symptoms (assessed via PHQ-8) were evaluated through a combination of surveys administered remotely and during in-person laboratory visits at intake, pre-intervention, and post-intervention. Both approaches exhibited high feasibility, producing significant pre–post improvements in dietary quality across the total sample (d = 1.93). Pre-post reductions in depressive symptoms were observed in the meal delivery condition (d = 1.62), but not the nutritional guidance condition (d = .54); these between-group differences should be interpreted cautiously given the pilot design and limited statistical power. Within-person analyses for the total sample indicated that greater improvements in dietary quality were associated with larger reductions in depressive symptoms, with dietary quality explaining 17% of the variance in depressive symptoms (semi-partial R² = .17). These findings provide insight into the potential of minimally processed dietary interventions for improving dietary quality and mental health and lay the groundwork for larger trials to assess longer-term effectiveness and sustainability.

Supplementary Information

The online version contains supplementary material available at 10.1038/s41598-026-45991-3.

Introduction

Major depression is one of the most common mental disorders in the United States, affecting nearly 1 in 6 adults at some point in their lives¹. Depression can significantly impact daily functioning, increase risk of suicidal thoughts and behaviors, and elevate risk of other chronic health conditions such as cardiovascular disease, cancer, and diabetes^2,3. Standard treatments for depression include psychotherapy and pharmacological therapies, which are often offered in combination for more severe cases of depression. While these treatments are generally effective, there are notable limitations. An estimated 30% of patients have been found to be treatment-resistant and nearly 40% of individuals with depression do not receive any treatment^4,5, often due to barriers such as lack of access, stigma surrounding mental health, and long waiting times for initial treatment consultations^6–8. These challenges have led to growing interest in integrative approaches for managing depression, particularly those that involve lifestyle modification⁹. Among these, dietary quality has emerged as a promising area of research, both as a risk factor and a potential component of treatment for depression.

High consumption of ultra-processed foods (i.e., foods that have been significantly changed from their natural state through industrial processing and often contain added sodium, sugar, and fats) has been consistently associated with greater depressive symptoms in numerous longitudinal cohort and systematic reviews^8,10–16. This body of work is complemented by meta-analytic evidence, suggesting that high ultra-processed food consumption is cross-sectionally associated with up to a 44% increased odds of having depressive symptoms compared to lower consumption^12,16. In contrast, prospective studies indicate that consuming a diet composed primarily of minimally processed foods (i.e., foods that have undergone minimal changes to their natural state such as fruits, vegetables, and fresh meats) is associated with lower risk of depressive symptoms or slower symptom development^17,18. Several randomized controlled trials of dietary counseling interventions that encouraged intake of minimally processed foods over ultra-processed foods found similar associations between improved dietary quality and reductions in depressive symptoms^19–23. Together, these findings suggest that dietary interventions prioritizing minimally processed foods may offer benefits for individuals with depression.

Importantly, most dietary interventions for depression have focused on counseling individuals to follow Mediterranean-style diets, which emphasize fruits, vegetables, whole grains, legumes, fish, poultry, nuts, and healthy fats, while limiting added sugars, refined carbohydrates, saturated fats, and processed meats. Although these approaches have been found to improve mental and physical health^24,25, the complexity of the diets themselves, such as knowing what and how much to eat, accessing ingredients, and adapting to personal or cultural preferences, can make adherence challenging²⁶. Moreover, the strategies used to support dietary change (e.g., counseling, educational materials, motivational interviewing, structured meal planning) still place substantial effort and cognitive load on participants, including tracking portions and following detailed rules^19–23. This is critical as the effort required to plan, shop for, and prepare minimally processed foods can be a significant obstacle to adhering to more minimally processed dietary recommendations, and a major factor contributing to a return to highly convenient ultra-processed food intake^27–29. This may be especially true for individuals experiencing depressive symptoms, including fatigue, low mood, and avolition. Developing lower-effort interventions that reduce ultra-processed food intake while minimizing cognitive and practical burdens may improve feasibility, accessibility, and long-term adherence for this clinical population.

The current study

Our study aims to address these challenges by directly providing convenient access to minimally processed foods through a commercial meal delivery service. Commercial meal and meal-kit delivery services have become increasingly popular due to their convenience, variety of food options, and affordability, and may offer an innovative solution to reduce challenges related to food planning and preparation^30–33. Additionally, preliminary evidence from meal delivery interventions targeting food insecurity and malnutrition in older adults suggests that delivery alone may confer mental health benefits^34–36, even when meals are not explicitly designed to emphasize minimally processed foods. Accordingly, in the current study, we combined these approaches to examine whether providing convenient access to minimally processed foods via meal delivery could reduce depressive symptoms to a greater effect than counseling individuals to self-implement dietary changes.

Our primary objective was to test the feasibility and acceptability of this novel combined approach through a pilot investigation of a two-week dietary intervention among individuals with moderate to moderately severe depressive symptoms, while providing preliminary estimates of short-term symptom change to inform a longer, fully powered efficacy trial. This two-week time frame was guided by the feasibility-focused aims and evidence from prior dietary and physical activity interventions indicating that depressive symptoms can shift within the first few weeks of behavior change^20,35,37. In addition, the broader depression treatment literature underscores the prognostic value of early symptom change, with a meta-analysis of antidepressant trials demonstrating that lack of improvement within the first two weeks strongly predicts a low likelihood of subsequent response or remission³⁸. Taken together, this evidence supports the use of a two-week pilot to evaluate feasibility and characterize early symptom trajectories, while recognizing that longer interventions and follow-up are required to determine sustainability and underlying mechanisms.

For the dietary intervention, participants were randomly assigned to one of two conditions: meal delivery or nutritional guidance. All participants were instructed to eat a diet composed primarily of minimally processed foods for two weeks and were provided with a packet of nutritional guidance on recommended foods and beverages to consume and to avoid. To support participants in following the nutritional guidance, the meal delivery condition also received three meals a day through a commercial meal delivery service. The nutritional guidance condition served as the active control, as participants were still responsible for planning, shopping for, and preparing their own meals in accordance with the nutritional guidance. Notably, though, this nutritional guidance condition expands on prior approaches by examining a minimally processed dietary approach distinct from the Mediterranean-style diet, which may be simpler and more adaptable to existing eating patterns.

We first evaluated whether there were differences in the feasibility and acceptability of the two intervention approaches through retention rates and quantitative debriefing responses. Feasibility was determined by retention across all study visits and participants’ self-reported success in following the nutritional guidance. Acceptability was evaluated through participants’ ratings of meal taste and fillingness (meal delivery only), perceived impact of the intervention on mental health, and intentions to continue to implement the nutritional guidance or use the meal delivery service after the intervention. We then evaluated preliminary effectiveness of each intervention for reducing depressive symptoms, as assessed by the Patient Health Questionnaire-8 (PHQ-8)³⁹. Specifically, we expected that both conditions would report reductions in depressive symptoms from pre- to post-intervention, but that the magnitude of reduction would be greater for the meal delivery condition than the nutritional guidance condition (H1). We also expected that greater improvements to dietary quality would be associated with greater reductions in depressive symptoms across both conditions and examined whether this within-person association was more pronounced in the meal delivery condition compared to the nutritional guidance condition (RQ1).

Materials and methods

Transparency and openness

This research was approved by the Institutional Review Board at the University of Michigan (HUM00231527) and the primary aims and outcome measures were preregistered on ClinicalTrials.gov on January 16, 2024 (NCT06242665). All study procedures were conducted in accordance with applicable institutional guidelines and regulations, and in accordance with the principles of the Declaration of Helsinki, including informed consent and protection of participant welfare. The present manuscript focuses on primary methods, feasibility, and preliminary dietary and depressive symptom outcomes (corresponding to Aim 2 on ClinicalTrials.gov). Analyses of other preregistered outcomes—including ecological momentary assessment (EMA), continuous glucose monitoring (CGM), and Fitbit-derived activity measures—will be reported in future publications. For brevity, only procedures and measures relevant to the present analyses are described below. The full study protocol, measures, data for the current aims, and analytic code are available on the Open Science Framework (OSF; https://osf.io/pcufr/).

Study design

This pilot study used a randomized two-group design to evaluate the feasibility, acceptability, and preliminary effectiveness of the commercial meal delivery service for reducing depressive symptoms, compared to self-implemented nutritional guidance. The active study period included a combination of remote activities and three in-person visits over the course of ~24 days, with a 1-week baseline occurring between the first and second visit, and the 2-week dietary intervention occurring between the second and third in-person visit (see Fig. 1 for study flow diagram). Study data were collected and managed using REDCap electronic data capture tools hosted at the University of Michigan, supported by the Michigan Institute for Clinical and Health Research (MICHR) [UM1TR004404].

Fig. 1.

Timeline and study flow of an average participant. Before coming into the lab, participants completed screening, were assessed for eligibility during the consent call, and then received a remote survey to complete before their Visit 1, which included the PHQ-8. On Day 0, participants came to the lab for Visit 1, during which they completed many baseline tasks, including their Visit 1 PHQ-8 (re-confirming eligibility) and Diet ID assessment. Participants completed a week of remote data collection (i.e., baseline period) from Days 1 to 7 while continuing to eat their typical diet. On Day 7, participants returned to the lab for Visit 2, during which they repeated baseline tasks, including their Visit 2 PHQ-8 and Diet ID assessment. At the end of Visit 2, participants received instructions for their randomized intervention condition: meal delivery or nutritional guidance. Participants completed two weeks of remote data collection (i.e., intervention period) from Days 8 to 21 while eating a minimally processed diet according to their intervention condition. On Day 21, participants returned to the lab for Visit 3, during which they repeated baseline tasks, including their Visit 3 PHQ-8 and Diet ID assessment, and completed a debriefing interview. Active study participation concluded upon completion of Visit 3. Participants were sent remote follow-up surveys at 1 and 6 months. Note that this figure includes variables pertinent to the current paper’s aims. Complete measures assessed at each timepoint are included in SM.

Sample size and power

As this was a pilot study, we aimed to recruit fifty participants (n > 30 in the meal delivery condition; Lancaster et al., 2004); however, inflation-driven rises in food costs caused funding to be exhausted sooner than expected, which led to an earlier conclusion of the recruitment process. Additionally, the partnering meal delivery company was liquidated days after the final participant completed the study, preventing further recruitment through alternative funding sources. Given the smaller-than-anticipated sample size (N = 31 participants; 20 meal delivery, 11 nutritional guidance) and because this is a pilot study, we conducted a sensitivity-style power analysis to estimate what size of group difference the study could realistically detect. This analysis was intended to provide context for interpreting the precision of the results rather than to justify the sample size. This estimate assumed a pooled standard deviation of change scores of 2.50, based on a between-subject standard deviation of 4 and an intraclass correlation (ICC) of 0.8 for repeated depressive symptom measurements. This ICC was selected based on a systematic review reporting pooled ICC estimates of 0.84 (95% CI: 0.81–0.88) for PHQ-9 scores⁴⁰, which closely correspond to PHQ-8 assessments. Using a simplified comparison of change scores (Visit 3 PHQ-8 score minus Visit 2 PHQ-8 scores), the study had 80% power to detect a large effect (Cohen’s d = 1.09) with group sizes of 20 and 11 at α = 0.05. Under these assumptions, the detectable between-group difference in PHQ-8 change from Visit 2 to Visit 3 was 2.76 points. Overall, this indicates that the study was not large enough to reliably detect small or moderate differences between groups, and that effectiveness findings should therefore be preliminary and hypothesis-generating.

Eligibility and informed consent

Participants were recruited from UMHealthResearch (https://umhealthresearch.org) and social media platforms from December 9, 2023 to July 28, 2024 to participate in a research study about how different foods impact feelings. Those interested in the study completed an online screening questionnaire on REDCap assessing eligibility criteria to ensure they could complete the dietary intervention and remote data collection tasks (e.g., EMA, CGM). Inclusion criteria were (a) 18 + years of age, (b) owning an Android or iPhone smartphone, (c) able to access phone every 90 min, (d) able to attend consent call over zoom/phone, (e) a score of at least 10 on the PHQ-8 assessed at consent, (f) endorsing daily intake of at least 2 ultra-processed foods the day before, (g) willing to come to three in-lab visits, (h) willing to log meals and follow dietary guidelines, (i) willing to share address with food delivery service, and (j) able to speak, write, read English fluently. Exclusion criteria were (a) regular smoking/vaping of nicotine, (b) history of medication that impacts reward or eating (Metformin, Lithium), (c) severe mental health conditions (e.g., schizophrenia, bipolar disorder), (d) certain medical conditions (e.g. diabetes, hypoglycemia), (e) endorsement of a suicide attempt or hospitalization for psychiatric reasons in last year, (f) diagnosis of restrictive eating disorder in last 5 years, (g) diagnosis of disorders that impact eating (e.g. cancer, hypothyroidism), (h) pregnant or post-partum, (i) frequent night shift work or irregular shifts, (j) severe allergies or food preferences not able to be accommodated, and (k) high score (i.e., > 3.4) on the Adult Picky Eating Questionnaire. Of note, the single-day ultra-processed food criterion was intended to confirm current exposure and ensure participants were not already following a minimally processed diet; usual dietary patterns were captured during Visit 1. If eligibility criteria were met, participants scheduled a consent call to confirm eligibility based on the PHQ-8, provide written informed consent, and schedule study visits.

Study protocol

A detailed study protocol, including a list of all measures, is included in Supplemental Material (SM1). After consenting to participate in the study, participants were emailed a link to a remote survey on REDCap to be completed prior to coming to the lab for their first in-person visit (M = 6.13 days prior, SD = 5.21, range = 0–16). This remote survey assessed demographic information and various individual characteristics related to eating motives, self-regulation, mental health history, and current depressive symptoms (PHQ-8).

Twenty-four hours before  in-lab Visit 1 (i.e., intake assessment), eligibility was re-confirmed using the PHQ-8 (hereon referred to as “Visit 1 PHQ-8”). If the Visit 1 PHQ-8 was ≥ 8 (lowered from the initial screening requirement of ≥ 10 to account for regression to the mean)⁴¹, participants were asked to come to the lab for their scheduled in-person Visit 1. This visit included a series of tasks, including completion of a dietary assessment (Diet ID) and a questionnaire battery, placement of a CGM sensor, provision of a Fitbit Charge 6, training on an EMA mobile app (e.g., LifeData) for remote data collection, and body composition measurements (e.g., height, weight, bioelectrical impedance analysis). Remote data collection occurred during the following week while participants continued to eat their typical diet (i.e., baseline period).

After approximately 7 days (M = 10.10, SD = 1.47, range = 7–12), participants returned to the lab for in-person Visit 2 (i.e., pre-intervention assessment). This visit included a similar series of tasks, including repetition of the dietary assessment and questionnaire battery (including the PHQ-8) from Visit 1, replacement of the CGM, and body composition measurements. After all study tasks were completed, participants received dietary instructions to be followed over the next two weeks in accordance with their randomly assigned food provision condition. Remote data collection occurred during these two weeks (i.e., intervention period).

After approximately 14 days (M = 16.68, SD = 1.87, range = 15–24), participants returned to the lab for their final in-person Visit 3 (i.e., post-intervention assessment), where they completed the same dietary assessment and questionnaire battery, body composition measurements, and a debriefing interview. One participant in the meal delivery condition had only 8 days of dietary intervention due to a food recall and subsequent delay in delivery services. Their Visit 3 was postponed by one week to ensure they completed a full week of dietary intervention and remote tasks. Analyses were conducted with and without this participant; findings did not change, and thus their data is included in the reported analyses. Participants were contacted 1 month and 6 months later to complete the same dietary assessment and questionnaire battery. The final follow-up assessment was completed on January 22, 2025.

Participants received $5 in compensation for completing the initial remote questionnaire battery, $25 for each in-lab visit, $27 for all EMA tasks and a $25 bonus if their EMA completion rate was above 80%, and $15 each for the 1-month and 6-month follow-ups. Participants were also able to keep the Fitbit Charge 6 device provided to them for the study.

Adverse event protocol

Participant safety was monitored through both structured and unstructured procedures, including repeated assessment of depressive and anxiety symptoms, disordered eating–related outcomes, and weight (to monitor for extreme weight loss). Participants also responded to open-ended questions in debriefing interviews at the end of each visit (e.g., “Do you feel like anything we asked you to do changed your eating behaviors, mood, or routine in the past week?”; “Is there anything else you think we need to know about your experiences in the study?”) and could report any other concerns via participant-initiated communication with the study team. Predefined escalation and discontinuation procedures were in place if safety concerns arose; none were identified through structured or unstructured monitoring.

Randomization

A computer-generated random number sequence in Microsoft Excel was used to allocate participants to the meal delivery or nutritional guidance condition using block randomization with a 2:1 allocation, stratified by sex (i.e., female, not female). This allocation favored a larger sample in the meal delivery group to allow more participants to receive the potentially optimized treatment and to collect richer data on experiences with the meal delivery service. This approach both improves the precision of estimates and helps identify factors that might influence outcomes, particularly within the constraints of a small pilot sample. The project coordinator generated the random number sequence, and participants were assigned a sequential study number upon completing their consent call. The coordinator also facilitated meal or snack provision (i.e., selecting foods and arranging delivery). To deliver condition-specific instructions, the researcher responsible for intervention delivery was unblinded prior to Visit 2. To separate intervention delivery from outcome assessment, a different researcher administered study tasks at Visit 3 and remained blinded until the final debriefing interview at the end of the study visit. To further minimize potential expectancy effects, all researchers were blinded to study hypotheses and primary analytic aims, and interactions with participants followed standardized protocols using word-for-word scripts. Dietary quality (Diet ID) and depressive symptom (PHQ-8) assessments were also self-reported by participants during study visits in a private room via online survey software.

At the end of Visit 2, all participants were instructed to consume a diet consisting primarily of minimally processed foods starting the next day until their Visit 3. Each participant was provided with a handout with nutritional guidance on how to eat in accordance with a minimally processed diet, which included descriptions and examples of recommended foods and beverages to eat and to avoid 1(Table 1), and a sample menu with ideas on how to structure one’s daily meal and snack intake in accordance with a minimally processed diet. A selection of customized meals or snacks, depending on condition, were delivered to participants’ home addresses prior to their Visit 2 (described below). Co-author K. Sonneville is a registered dietitian who oversaw the development of the nutritional guidance and study instructions, selection of the commercial meal delivery service, and identification of meal and snack options to ensure alignment with a minimally processed dietary pattern.

Table 1.

Examples of minimally processed foods and beverages recommended during the intervention and highly processed foods and beverages participants were asked to avoid (full descriptions provided to participants with examples are included in SM1).

Food Category	Highly processed foods	Minimally processed foods
Recommendation	Not Recommended	Recommended
Description provided to participants	Foods that have been manufactured from ingredients that do not occur in their natural state; ingredients often include added fats, added sugars, refined carbohydrates, preservatives, and artificial sweeteners and flavors	Foods that are purchased and eaten closer to their natural state
Examples	– Foods with added sauces or syrups that contain preservatives and/or added sugar – Added sugar (e.g., sugar, maple syrup, honey, corn syrup, cane juice, malt syrup/sugar, inverted sugar syrup, molasses, sucrose, fruit juice concentrates) – Artificial sweeteners (e.g. aspartame, sucralose, saccharin, stevia leaf extracts, monk fruit, acesulfame potassium, xylitol, erythritol) – Breaded or fried foods – Cereals, breads, pastas, or crackers made with refined grains and/or including added sugar – Baked goods, cookies, candy, ice cream, or other frozen desserts – Combination frozen foods, such as pizza, burritos, or frozen dinners – Juices, sodas, or energy drinks	– Fresh vegetables and fruits – Frozen or dried vegetables and fruits packaged without added sugars – Fresh or frozen meat, fish, and poultry without added sauces or breading – Fresh eggs – Milk, unless flavored or sweetened – Beans and legumes, packaged fresh, dry, or frozen – Nuts and seeds, packaged without added sugars or flavors – Whole, intact grains – Herbs and spices, fresh or dried – Plain, unsweetened yogurt – Butter or oil added when preparing foods – Tea or herbal tea, packaged as tea bags or loose-leaf tea – Coffee, packaged as whole coffee beans or ground

The aim of the dietary interventions was to improve dietary quality, rather than produce weight loss or symptoms of withdrawal (e.g., alcohol, caffeine). Thus, all participants were encouraged to supplement study-provided food with minimally processed foods ad libitum. They were also instructed to sustain their regular alcohol and caffeine intake and were provided with customized suggestions on how to modify beverages in accordance with the nutritional guidance (e.g., remove added sweeteners, replace mixers). A full description of this protocol with copies of the nutritional guidance are included in Supplemental Material (SM1) and on the OSF project page (https://osf.io/pcufr/).

Meal delivery condition

Participants assigned to the meal delivery condition received a tailored set of instructions explaining that, to help them follow the nutritional guidance, they would be sent all of their meals for the next two weeks to their home address from a commercial meal delivery service company.

The company that the study team partnered with was a healthy pre-made meal delivery subscription service that catered to a variety of eating plans and was known for its use of high-quality whole foods (the company representatives have asked for the company name to be omitted from publication). Specifically, participants were provided with three meals per day (breakfast, lunch, and dinner) for a total of 21 meals per week. The meals to be delivered were selected by the project coordinator on the meal delivery service’s website and were required to be in accordance with a Mediterranean, Paleolithic, or Whole 30 diet; three diets that primarily consist of minimally processed ingredients with low amounts of added sugars and refined carbohydrates^42–45. Meals selected were also informed by allergies and food preferences indicated by the participant during their consent call.

Selections of meals provided to each participant in the meal delivery condition and estimated nutritional composition of each meal is included in Supplemental Material (SM1 and SM2, respectively). Nutritional information for these meals was not available directly from the meal delivery company due to its liquidation. Nutritional values were therefore obtained from the MyNetDiary food database (https://www.mynetdiary.com/food-database.html) and should be interpreted as approximate estimates rather than verified nutrient composition. Mean (SD) values across meals were as follows: saturated fat 6.12 g (3.05), sodium 534.35 mg (148.37), added sugar 0.48 g (1.79), dietary fiber 5.55 g (3.19), and protein 25.21 g (7.81). These estimates suggest that the meals were generally consistent with a minimally processed dietary pattern, with lower added sugar and higher dietary fiber being particularly indicative of foods that are less processed and more nutritionally dense⁴⁹.

Participants were instructed to do their best to eat the meals provided to them, but to follow the provided nutritional guidance if they were in a situation where they needed to eat something different. Participants were also recommended to eat three meals and at least one snack each day, but to prioritize their normal eating pattern and incorporate the nutritional guidance if it was different (e.g., 2 meals and 3 snacks each day). Finally, the researcher and participant engaged in a planning exercise to prospectively identify situations that might make following the nutritional guidelines more challenging, and to create contingency plans to manage those situations.

Nutritional guidance condition

Participants assigned to the nutritional guidance condition received a tailored set of instructions explaining that a small selection of snacks had been sent to their home address. The snacks included 2–3 minimally processed items (e.g., dried fruit, nuts) that were high in fiber and/or protein, had no added sugar, and totaled less than $20. These items were intended to serve as illustrative examples of minimally processed options and to build rapport, rather than as a prescribed, nutrient-controlled, or adherence-supporting component of the intervention. Snack selections were made by the project coordinator and were informed by participant-reported allergies and food preferences collected during the consent call. A detailed description of personalized snack selections with portion sizes for each participant is included in Supplemental Material (SM1). Similar to the meal delivery condition, participants were instructed to follow the nutritional guidance for their main meals and were recommended to eat three meals and at least one snack each day (prioritizing their normal eating pattern if it was different). Finally, the researcher and participant engaged in the same planning exercise to prospectively identify challenging situations and to create contingency plans to manage them.

Measures

Dietary quality

Dietary quality was assessed with Diet ID™–a commercial platform (www.DietID.com) that provides a rapid assessment of dietary intake through a computer choice task⁴⁶. Participants are presented with pairs of composite images representing established dietary patterns in North America and select the image that best reflects their current diet. The program refines the selections by providing new pairs of images until the best possible fit is achieved. Based on participants’ responses, Diet ID provides estimates of energy intake, macronutrients, micronutrients, and overall dietary quality. Nutrient estimates from this method have been validated against standard dietary assessment approaches, including 24-h dietary recalls and food frequency questionnaires^46–48. Importantly, Diet ID allows for the assessment of overall dietary quality without introducing additional demands or behavior-change strategies (e.g., detailed food logging or self-monitoring) that could alter eating behavior, dietary adherence, or study burden⁴⁸.

In the present study, Diet ID–derived estimates were used to examine changes in dietary quality as well as exploratory changes in added sugar and dietary fiber intake, two nutrient components indicative of ultra-processed food consumption⁴⁹. The Diet ID dietary quality score is a validated 1–10 metric derived from image-based dietary pattern assessment using an algorithm informed by the Healthy Eating Index (HEI) 2020. The HEI evaluates alignment with the Dietary Guidelines for Americans, 2020–2025 edition, based on 13 dietary components⁵⁰. Diet ID applies HEI-based analysis to classify overall dietary patterns and then derives a 1–10 diet quality score that reflects alignment with federal dietary guidance while accounting for variation in intake patterns across diet types. Higher scores indicate greater alignment with these guidelines, which emphasize adequate consumption of fruits, vegetables, whole grains, dairy, and protein foods and moderation of refined grains, added sugars, and sodium; an overall pattern consistent with a minimally processed diet⁵¹.

In this study, participants were asked to complete the Diet ID task twice per in-person visit, once for their typical weekday diet and once for their typical weekend diet during the previous week. A weighted average [(weekday dietary quality*5 + weekend dietary quality*2)/7] was calculated for each assessment (i.e., Visit 1, Visit 2, Visit 3) for analysis.

Depressive symptoms

Depressive symptoms were assessed during the remote assessment and each in-person visit using the PHQ-8, a brief measure of depression severity³⁹. Participants score each of the first 8 DSM-IV criteria (e.g., “Little interest or pleasure in doing things”) on a 4-point scale from 0 (not at all) to 3 (nearly every day) based on how they felt over the past two weeks. Responses to each item are summed for a composite score: the final scores of 0–4, 5–9, 10–14, 15–19, 20–24 are the ranges for none, mild, moderate, moderately severe and severe, respectively.

Body mass index (BMI)

BMI was examined both as a potential covariate to ensure that intervention effects on depressive symptoms were not influenced by any changes in weight across the study, and as a general indicator of energy balance as substantial weight change would suggest meaningful shifts in energy intake across the study period. Height was measured to the nearest tenth of a centimeter at Visit 1 using a seca 213 portable stadiometer. Weight and body composition were measured at each in-person visit using bioelectrical impedance analysis (InBody 570). BMI was calculated for each assessment (i.e., Visit 1, Visit 2, Visit 3) using the standard formula (Weight (lb)/Height (in)² × 703).

Intervention feedback

At the end of Visit 3, several quantitative debriefing items were asked by a researcher through an in-person interview to determine if there were differences in the feasibility and acceptability of each intervention (i.e., meal delivery, nutritional guidance). Adherence to nutritional guidance was assessed through one item: on a scale of 0 to 100 percent, with 0 being not at all successful, and 100 being completely successful, how successful was the participant at following the nutritional guidance we provided. Diet similarity was assessed through one item: on a scale from 0 (not at all) to 100 percent (completely the same), how similar the nutritional guidance (nutritional guidance condition) or the delivered meals (meal delivery condition) were to the participant’s regular diet. Participants in the meal delivery condition were also asked to report intervention-specific items, including how tasty the provided meals were (overall) on a scale from 1 (very bad) to 7 (very good), and how filling the provided meals were (overall) on a scale from 1 (not filling at all) to 7 (very filling). Participants in the nutritional guidance condition were not asked these questions, as they had the autonomy to self-implement the nutritional guidance according to their personal preferences for type and amount of food.

To assess perceived changes to mental health, participants reported on a scale from 1 (made it much worse) to 7 (made it much better) what impact the provided nutritional guidance (nutritional guidance condition) or meal delivery service (meal delivery condition) had on their mental health. Finally, to assess intentions to continue, participants reported on a scale from 1 (not at all) to 7 (definitely) if they planned to keep trying to follow the nutritional guidance (nutritional guidance condition) or meal delivery service (meal delivery condition) after the study.

Data analytic plan

Preliminary analyses

We first constructed a CONSORT diagram detailing the flow of participants assessed for inclusion, randomized, excluded after randomization, and retained through follow-up analyses⁵². Data were then assessed for normality and outliers. Independent sample t-tests and chi-square analyses were conducted to compare participant characteristics based on condition in age, sex, gender, race, ethnicity, education, income, and initial assessments of BMI (Visit 1), dietary quality (Visit 1), and depressive symptoms (remote assessment).

For the remaining analyses, linear mixed-effects models were used to evaluate change over time while accounting for within-person correlations across repeated assessments via a random intercept for participant. Linear mixed-models include all available data under standard missing-at-random assumptions without requiring imputation. All models included fixed effects of condition, visit (categorical), and the condition × visit interaction and were fit using the lme4 package in R⁵³. Planned contrasts then tested (a) stability across the pre-intervention assessments and (b) change from pre- to post-intervention (i.e., Visit 2 to Visit 3). Because multiple pre-intervention assessments were collected and diet composition and depressive symptoms can fluctuate over short intervals, a summary contrast tested change from Visit 3 to the average of the pre-intervention assessments to provide a more stable estimate of overall pre- to post-intervention change and to reduce the influence of short-term variability. For the dietary quality analysis, the pre-intervention average included Visit 1 and Visit 2; for the depressive symptom analysis, the average included the remote assessment Visit 1, and Visit 2. Estimated marginal means were used to obtain these contrasts.

Feasibility and acceptability

In addition to study retention, feasibility was evaluated at two levels: (1) whether each intervention approach produced improvements in dietary quality and (2) whether participants reported successful implementation and positive perceptions of the intervention experience.

Dietary quality was evaluated as the primary indicator of change during the intervention. Linear-mixed models were also conducted for estimated caloric intake, added sugar, and dietary fiber derived from Diet ID to ensure that participants in both conditions maintained energy intake across the study period and to examine whether improvements in dietary quality were accompanied by changes in dietary components associated with ultra-processed food consumption (i.e., higher dietary fiber, lower added sugar)⁴⁹. Changes in BMI were also evaluated as an additional indicator of energy balance.

Participant-level feasibility and acceptability were evaluated using quantitative debriefing measures administered at the end of Visit 3. Descriptive analyses were first conducted to characterize participants’ experiences with each intervention approach. Correlation analyses were then used to examine associations among feasibility-related perceptions (e.g., whether liking of the food was associated with intentions to continue the intervention).

Preliminary effectiveness

We next examined preliminary evidence for changes in PHQ-8 scores from pre- to post- intervention, and whether reductions were greater for the meal delivery condition than the nutritional guidance condition (H1). Because this pilot study was not powered to detect small or moderate between-group effects, analyses of depressive symptom change were intended to provide preliminary, hypothesis-generating estimates rather than definitive tests of efficacy.

To directly evaluate whether changes in dietary quality were associated with changes in depressive symptoms, we also examined the magnitude of the within-person association between dietary quality and PHQ-8 scores across Visits 1–3 (Diet ID was not collected at the remote assessment). This analysis used a linear mixed-effects model predicting PHQ-8 scores from dietary quality, with an interaction term between dietary quality and intervention condition to test whether this association was stronger in the meal delivery condition compared to the nutritional guidance condition (RQ1). Note that participants were excluded if their Visit 1 PHQ-8 score was less than 8; thus, comparisons of changes involving Visit 1 PHQ-8 scores may include regression-to-the-mean effects.

Results

Study participants and retention

Thirty-one participants were enrolled in the study and completed all in-person and remote study tasks (see Fig. 2 for CONSORT diagram). Twenty participants were randomized to the meal delivery condition, and 11 were randomized to the nutritional guidance condition. Demographics and descriptives for each condition and the total sample, including age, sex, race, ethnicity, education, income, BMI, dietary quality, and PHQ-8 scores are included in Table 2. Notably, all participants completed all study assessments, suggesting high retention and preliminary feasibility of both intervention approaches.

Fig. 2.

CONSORT diagram. Note: In the original protocol, participants completed the PHQ-8 upon arrival for Visit 1 to ensure they still reported a score ≥ 8 . Three of six participants were excluded based on this criterion. Thus, to reduce participant burden (i.e., unnecessary commute to the lab), the protocol was changed to assess the PHQ-8 through an online survey that was emailed 24 hours before one’s scheduled Visit 1. Nine participants were excluded in the latter protocol. For brevity, these participants are aggregated as ineligible per pre-Visit 1 PHQ-8 determination in the CONSORT diagram

Table 2.

Demographic and descriptive characteristics for each condition and the total sample.

	Nutritional guidance	Meal delivery	Total sample
	(n = 11)	(n = 20)	(N = 31)
Sex assigned at birth
Female	9 (81.8%)	16 (80.0%)	25 (80.6%)
Male	2 (18.2%)	4 (20.0%)	6 (19.4%)
Race*
Asian	3 (27.3%)	3 (15.0%)	6 (19.4%)
Black or African-American	1 (9.1%)	0 (0.0%)	1 (3.2%)
White	6 (54.5%)	18 (90.0%)	24 (77.4%)
Other	1 (9.1%)	0 (0.0%)	1 (3.2%)
Ethnicity
Hispanic or Latino	0 (0.0%)	2 (10.0%)	2 (6.5%)
Not Hispanic or Latino	11 (100%)	18 (90.0%)	29 (93.5%)
Education
Some college	1 (9.1%)	3 (15.0%)	4 (12.9%)
Associates degree	0 (0.0%)	2 (10.0%)	2 (6.5%)
Bachelor’s degree	6 (54.5%)	9 (45.0%)	15 (48.4%)
Advanced degree	4 (36.4%)	6 (30.0%)	10 (32.3%)
Income
< $50,000	2 (18.2%)	7 (35.0%)	9 (29.0%)
$50,000—$99,999	1 (9.1%)	5 (25.0%)	6 (19.4%)
> $100,000	8 (72.8%)	7 (35.0%)	15 (48.4%)
Prefer not to answer	0 (0.0%)	0 (5.0%)	1 (3.2%)
Age (SD)	39.27 (14.16)	38.80 (12.43)	38.97 (12.83)
BMI (SD)
Visit 1	27.30 (5.73)	28.70 (5.11)	28.20 (5.29)
Visit 2	27.40 (5.88)	28.40 (5.09)	28.00 (5.32)
Visit 3	27.00 (5.69)	28.30 (4.85)	27.90 (5.11)
Depressive symptoms (SD)
Remote	13.09 (4.16)	13.95 (3.03)	13.65 (3.43)
Visit 1	14.36 (4.84)	14.45 (2.93)	14.42 (3.64)
Visit 2	11.55 (3.47)	13.45 (4.11)	12.77 (3.95)
Visit 3	10.18 (3.22)	9.40 (4.57)	9.68 (4.10)
Dietary quality (SD)
Visit 1	5.16 (2.33)	4.72 (2.06)	4.88 (2.13)
Visit 2	5.45 (2.40)	4.30 (1.61)	4.71 (1.97)
Visit 3	8.21 (7.95)	7.95 (2.52)	8.05 (2.19)

*Participants could select all that applied for race; may total > 100%. Depressive symptoms are represented by PHQ-8 scores, which may be interpreted in terms of severity: none (0–4), mild (5–9), moderate (10–14), moderately severe (15–19), and severe (20–24). Dietary quality refers to conformance with federal dietary guidance on a scale from 1 (low quality diet) to 10 (high quality diet); for example, a score of 4.88 indicates ~50% alignment with federal dietary guidance.

No continuous variables (e.g., age, BMI, PHQ-8 scores) showed significant departures from being normally distributed and no outliers were detected. At the time of their participation, 12 participants were receiving a combination of medication and therapy for depression (meal delivery = 45.0%, nutritional guidance = 27.2%), 9 were receiving medication only for depression (meal delivery = 25.0%, nutritional guidance = 36.4%), 3 were receiving therapy only for depression (meal delivery = 5.0%, nutritional guidance = 18.2%), and 7 were receiving no form of standard treatment for depression (meal delivery = 25.0%, nutritional guidance = 18.2%). One participant in the meal delivery condition was missing dietary quality at Visit 3 (post-intervention) due to researcher error (did not administer). There were no significant differences based on condition for any baseline variable (SM Tables A and B). A marginally higher proportion of participants identified as being White in the meal delivery condition (90.0%) than the nutritional guidance condition (54.5%). Thus, sensitivity analyses were conducted including race (coded as White vs. non-White to avoid sparse cells) as a covariate in all outcome models (SM Tables H & O). Results were unchanged when models additionally adjusted for race.

Feasibility of intervention approaches to improve dietary quality

Dietary quality at each assessment by condition and for the total sample is described in Table 2 and Fig. 3 (individual trajectories illustrated in SM Figures A and B). Participants reported mean dietary quality scores of 4.88 at Visit 1 and 4.77 at Visit 2, indicating they were only meeting about half of the food group and nutrient recommendations considered to comprise a healthy dietary pattern in the U.S. before the start of the dietary intervention⁵¹. Pairwise comparisons, described in Table 3, indicated that dietary quality did not significantly differ between the two pre-intervention assessments, indicating stability during the baseline period. Dietary quality increased significantly from Visit 2 to Visit 3, reaching an average of 8.05 and representing an overall improvement from approximately 50% to 80% alignment with dietary guidelines across conditions. The summary contrast comparing Visit 3 to the average of the pre-intervention assessments was also significant. There was no significant main effect of condition and no significant visit × condition interactions (SM Table C).

Fig. 3.

Improvement in mean dietary quality over time based on condition.

Table 3.

Planned pairwise contrasts of estimated marginal means by Visit for dietary quality and PHQ-8 scores.

Contrast	Intervention period	Estimate [95% CI]	p	d
Diet quality
Visit 1 – Visit 2	Baseline stability	.06 [-.99, 1.11]	.99	.04
Visit 2 – Visit 3	Intervention effect	−3.17 [−4.23, −2.11]	<.001	−1.93
Pre-avg – Visit 3	Intervention effect	−3.14 [−3.90, −2.38]	<.001	−1.91
PHQ-8 scores
Remote – Visit 1	Baseline stability	-.89 [−2.63,.86]	.54	-.35
Remote – Visit 2	Baseline stability	1.02 [.72, 2.76]	.42	.41
Visit 1 – Visit 2	Baseline stability	1.91 [.17, 3.65]	.03	.76
Visit 2 – Visit 3*	Intervention effect	2.71 [.96, 4.45]	<.001	1.08
Pre-avg – Visit 3	Intervention effect	3.68 [2.60, 4.76]	<.001	1.47

*Note: effect qualified by a visit × condition interaction.

In addition to dietary quality, estimated added sugar decreased (SM Table D) and estimated dietary fiber increased (SM Table E) across conditions, consistent with what would be expected for shifting to a more minimally processed diet⁴⁹. There was also no change in estimated calories (SM Table F) or BMI (SM Table G) over time in either condition, suggesting that the interventions improved the nutritional quality of participants’ diets without altering overall energy intake.

Participant-level feasibility and acceptability of each intervention approach

Distributions of responses to each of the debriefing questions are illustrated in SM Figures C – G. Within the nutritional guidance condition, participants generally reported high success at following the nutritional guidance (M = 86.82, SD = 9.02). On average, they reported their diets to be about 50% similar to their regular pre-intervention diets (M = 52.27, SD = 18.89). Participants also rated the effect of the nutritional guidance on their mental health around the scale midpoint (M = 4.64, SD = 0.67) and reported that they somewhat planned to continue to try to follow the nutritional guidance after the study (M = 5.00, SD = 0.77). There were no significant correlations among success, diet similarity, mental health impact, or intentions to continue.

Within the meal delivery condition, participants also generally reported high success at following the nutritional guidance (M = 79.80, SD = 18.57). The provided meals were rated as approximately 30% similar to their regular pre-intervention diet (M = 30.45, SD = 17.91) and received average ratings of tastiness (M = 4.25, SD = 1.74) and fillingness (M = 4.80, SD = 1.74). Participants reported that the effect of the meal delivery on their mental health was around the scale midpoint (M = 4.35, SD = 1.57), and that they did not have plans to continue the meal delivery service after the study (M = 2.05, SD = 1.43). Correlation analyses (SM Table M) indicated that each of these beliefs were strongly associated with perceived tastiness of the meals, such that participants who rated the meals as more tasty had stronger beliefs that that the meal delivery improved their mental health (r = 0.68, p < 0.001) and had stronger plans to continue the meal delivery service after the study (r = 0.48, p = 0.03). Additionally, participants who rated the meals as more similar to their regular pre-intervention diet reported higher success at following the nutritional guidance (r = 0.52, p = 0.02).

Preliminary effectiveness: post-intervention reductions in depressive symptoms

Average PHQ-8 scores at each assessment by condition and for the total sample are presented in Table 2 and Fig. 4 (individual trajectories are illustrated in SM Figures H and I). Participants reported mean PHQ-8 scores of 13.65 at the remote assessment, 14.42 at Visit 1, and 12.77 at Visit 2, indicating that depressive symptoms were generally consistent with the moderate to moderately severe range required for study inclusion. Pairwise comparisons, described in Table 3, indicated that PHQ-8 scores did not significantly differ between the remote assessment and Visit 1 or the remote assessment and Visit 2. However, PHQ-8 scores declined modestly between Visit 1 and Visit 2, suggesting some short-term fluctuation in symptoms prior to intervention initiation.

Fig. 4.

Reduction in mean PHQ-8 score over time based on condition. The bars represent ± 1 SE.

Following the intervention, PHQ-8 scores were significantly lower at Visit 3 compared to both Visit 2 and the average of the pre-intervention assessments. Change from Visit 2 to Visit 3 was moderated by condition (B = −2.67, p = 0.046; full model described in SM Table N) such that PHQ-8 scores significantly declined in the meal delivery condition (estimate = 4.05, 95% C.I. [1.98, 6.13], p < 0.01, d = 1.62) but not in the nutritional guidance condition (estimate = 1.36, 95% C.I. [−1.44, 4.16], p = 0.58, d = 0.54).

These between-group differences should be interpreted cautiously given the pilot design and limited statistical power. There was no significant main effect of condition and no other visit × condition interactions.

Preliminary effectiveness: within-person association between dietary quality and depressive symptoms

As described in Table 4, dietary quality was significantly associated with reductions in PHQ-8 scores, such that each unit increase in dietary quality was associated with a 0.87 decrease in PHQ-8 score after accounting for the random variability among participants. The strength of this association was not moderated by intervention condition.

Table 4.

Linear-mixed model predicting PHQ-8 score based dietary quality, intervention condition (coded 1 = meal delivery, 0 = nutritional guidance), and the interaction between dietary quality and intervention condition.

Within-person association
Predictors	Estimates		std. Error	95% CI	p	semipartial R2
(Intercept)	17.40		1.22	[14.97, 19.82]	<.001
Condition	−1.80		2.21	[−6.20, 2.59]	.42	.01
Dietary quality	-.87		.17	[−1.22, -.53]	<.001	.17
Dietary quality × condition	.31		.30	[-.29,.91]	.31	.01
Random effects
σ2			8.50
τ00 ID			8.09
ICC			.49
N ID			31
Observations			92
Marginal R2/conditional R2			.20/.59

Discussion

Previous intervention studies have found that dietary shifts toward more minimally processed foods are associated with reductions in depressive symptoms^19–23. Yet, the effort required to implement a minimally processed diet might undermine long-term dietary adherence^27,28, particularly for those with depression who may regularly experience fatigue, low mood, or avolition. This pilot study is among the first to evaluate the feasibility of using a commercial meal delivery service to increase the convenience of eating a minimally processed diet for individuals experiencing moderate to moderately severe depressive symptoms. Using a two-group experimental design, participants received nutritional guidance on how to implement a minimally processed diet, which was either fully facilitated by a commercial meal delivery service or self-guided. Findings support the feasibility of both approaches and provide preliminary evidence for their effectiveness to improve mental health outcomes. Important limitations and considerations for future research are described in detail below.

Feasibility and acceptability of dietary intervention strategies

Feasibility was evaluated at multiple levels, including study retention, participants’ ability to implement each intervention approach, and whether each approach produced meaningful improvements in dietary quality over the intervention period. The study’s zero drop-out rate, a key marker of feasibility and participant engagement, is particularly noteworthy given the common challenge of poor retention in studies involving participants with depression^21,54. Participants in both conditions also reported high success in following the nutritional guidance, and these self-reports were consistent with significant improvements in dietary quality from pre- to post-intervention, increasing from approximately 50% to 80% alignment with the Dietary Guidelines for Americans, 2020–2025 edition.

These findings suggest that both intervention approaches were capable of supporting short-term dietary change, an essential prerequisite for evaluating downstream mental health outcomes. However, it is important to note that dietary quality was assessed using Diet ID, a validated but self-report, image-based measure of overall dietary patterns^46–48. Although appropriate for capturing general patterns in this feasibility-focused pilot, Diet ID does not provide precise information on energy intake, macronutrient composition, or adherence to study-provided foods, limiting mechanistic inference and replication in studies targeting specific nutritional components. As a self-report measure, Diet ID may also be influenced by demand characteristics or other reporting biases leading participants to report healthier intake than actually consumed.

Relatedly, intervention fidelity relied on low-burden self-reported success rather than objective verification of meal consumption or detailed adherence monitoring. As a result, improvements in dietary quality cannot be interpreted as definitive evidence that study-provided meals were consumed as intended or that nutritional guidance was implemented with complete adherence. This approach was intentional: as a feasibility-focused pilot conducted in a population at elevated risk for disordered eating⁵⁵, we prioritized minimizing participant burden and reducing the potential negative effects of intensive dietary tracking on mood and disordered-eating symptoms^56–58. Future fully powered trials need to include strategies that more rigorously assess intervention fidelity while balancing participant burden and safety (e.g., passive physiological indicators where feasible, 24-h dietary recalls).

Preliminary effectiveness in reducing depressive symptoms

Significant reductions in depressive symptoms were observed from pre- to post-intervention for participants who received meal delivery, but not among those who self-implemented dietary change (aligning with H1). However, this finding should be interpreted cautiously given the modest sample size and limited statistical power. Importantly, depressive symptoms were significantly lower at post-intervention compared to the average of all pre-intervention assessments across conditions, suggesting that both intervention approaches may have contributed to symptom improvement across the study period. Because symptoms also declined between Visit 1 and Visit 2, prior to intervention initiation, it is difficult to fully disentangle intervention-related change from regression to the mean, baseline variability, or other study-related factors. These findings underscore the need for future trials with larger samples, aligned baseline and intervention durations, and a true control group (vs. active control). Such studies would help verify both inferences about causality and the magnitude of effects by isolating the specific contributions of meal delivery and dietary changes from other factors that might influence depression, such as participant demand characteristics, regression-to-the-mean, and other study-related tasks^41,59.

Given that prior self-guided dietary interventions have found antidepressant effects following shifts toward more minimally processed diets^19–23, it is also possible that meal delivery may confer more rapid benefits than self-guided dietary change. The mechanisms most often posited to underlie associations between diet quality and depression are primarily physiological. Diets high in ultra-processed foods are typically characterized by higher added sugars and lower dietary fiber and micronutrient density; patterns that have been implicated in processes linked to mood disturbance, including glycemic dysregulation, alterations in gut microbiota composition, increased systemic inflammation, and changes in neuroendocrine and neural signaling^16,60,61. Improvements in dietary quality may therefore contribute to reductions in depressive symptoms through gradual regulation of these biological systems (see Ekinci & Sanlier, 2023 for brief review)⁶². However, such physiological adaptations may require sustained dietary change, potentially longer than the two-week intervention evaluated in the present pilot.

In contrast, psychological and behavioral mechanisms related to the mode of implementation may operate more rapidly. The direct provision of meals may reduce cognitive and practical burdens related to food planning, shopping, and preparation—demands that may feel particularly overwhelming during depressive episodes and exacerbate perceived effort costs that contribute to low motivation, apathy, and avoidance. By reducing daily decision-making demands and self-regulatory load, meal delivery may facilitate adherence to dietary change and produce near-term improvements in functioning and mood^63–65, even before longer-term physiological processes have time to meaningfully shift. From this perspective, meal delivery may influence both dietary quality and depressive symptoms through immediate psychological processes, whereas self-guided dietary change may rely more heavily on sustained behavioral and biological processes. Longer-duration trials are needed to determine whether physiological adaptations emerge over time and whether facilitated meal provision confers incremental or accelerated benefits relative to self-guided dietary change.

Implications for dietary interventions as treatments for depression

Within individuals, greater improvements in dietary quality were associated with larger reductions in depressive symptoms, with dietary quality explaining approximately 17% of the variance. These findings suggest that short-term improvements in dietary quality may be associated with early reductions in depressive symptoms, consistent with prior lifestyle intervention research demonstrating measurable mood changes within the first few weeks of behavior change^20,35,37. Although the present pilot cannot establish causality, the observed associations align with evidence from broader literature positioning dietary quality as a promising modifiable contributor to mental health^19–23. This has broad implications in the United States where the majority of the American diet consists of ultra-processed food, which may be contributing to the rising prevalence of depression^8,10–16. Emphasizing the consumption of minimally processed foods could be a valuable yet underutilized strategy for addressing this issue^17,18.

There are various avenues through which dietary-focused interventions might support depression treatment. Dietary-focused interventions could be integrated into comprehensive depression treatment plans, either as stand-alone strategies or alongside therapy or medication. In this pilot sample, 80% of participants were already receiving therapy, medication, or both, yet still exhibited moderate to moderately severe depressive symptoms, suggesting that dietary approaches may offer adjunctive benefits. Although the present study was not powered to evaluate clinical effectiveness or interactions with ongoing treatment, future adequately powered trials should examine these possibilities directly. Taken together with prior research, the findings highlight dietary quality as a potentially overlooked factor in depression treatment that could complement standard care, particularly given the high prevalence of nutrient-poor, ultra-processed diets in the general population (> 60%)⁶⁶. Additionally, improved dietary quality may help to manage comorbid chronic health conditions, including cardiovascular disease, diabetes, chronic pain, and neurological disorders, which might otherwise worsen depression and make it more challenging to treat⁶⁷.

Prioritizing minimally processed foods in one’s diet may also serve as a beneficial approach for the many individuals who do not receive mental health care, which is common in the United States⁸. For instance, an estimated 21 million adults in the United States experienced at least one major depressive disorder in 2021, yet only 61.0% received treatment⁵. Some people may choose not to seek treatment due to stigma or beliefs around mental health, while others face a wide variety of barriers to access including finances, time for appointments, or lack of knowledge about treatment options⁶. Further, waiting times for initial treatment consultations have been found to average from 6 to 13 weeks, with some clinics seeing patients over a year after first referral^7,8. Delay in care can be detrimental for treatment outcomes, including decreased patient stability, and increased risk of relapse, crisis-related hospitalizations, and attempted and completed suicide⁶⁸. Evidence-based dietary interventions could address the crucial need for low-burden, affordable self-care strategies that provide a bridge between the onset of symptoms and time of first care⁶⁹. It is also plausible that some individuals may find dietary changes to be a less stigmatizing and more accessible entry point than traditional treatment options. A better understanding of how best to implement, personalize, and sustain these approaches in the context of mental health care are necessary next steps.

Optimizing dietary intervention strategies for depression

Participants’ responses to condition-specific debriefing questions provide initial insight into how to optimize meal delivery interventions to better meet the needs and preferences of individuals with depression. For example, participants in the meal delivery condition rated their meals as only 30% similar to their usual diet, suggesting limited compatibility with their existing eating patterns. The tastiness of the meals also appears to have been particularly relevant for participants’ perceptions of the intervention’s effectiveness on their mental health, as well as their intentions to continue meal delivery after the study. Together, these experiences highlight the importance of allowing participants autonomy in meal selection, including choices that align with taste preferences, rather than relying on pre-selected options by the study team. Such autonomy in meal selection may enhance perceived acceptability and enjoyment of meal delivery services and, in turn, support dietary adherence and mental health⁷⁰. It is also worthwhile to note that low intentions to continue the meal delivery service after the study may reflect the cost of the service, which was covered during the study, rather than a lack of interest in or enjoyment of the service itself. Future work should explore cost-effective models to better support long-term implementation, similar to how some health insurance plans or employer wellness programs subsidize gym memberships or nutrition programs.

In contrast, the simplified nutritional guidance provided in this study, which focused mainly on replacing ultra-processed and packaged foods with minimally processed options, may offer greater flexibility and adaptability to individual preferences and cultural needs. Allowing participants to select minimally processed food options within their existing eating patterns may also require fewer changes, which may potentially facilitate implementation and reduce cognitive load associated with preparatory tasks. This higher compatibility is supported by debriefing ratings from participants in the nutritional guidance condition, who indicated that the nutritional guidance was about 50% similar to what they were eating before the dietary intervention.

Finally, although the study was not powered to definitively test whether meal delivery yields greater improvements in depressive symptoms than nutritional guidance alone, this is one the first studies to examine commercially available meal delivery as a targeted intervention for depression. Further work is needed to evaluate the sustainability of both intervention approaches (i.e., meal delivery versus nutritional guidance) over time, especially as chronic and recurring depression may hinder one’s ability to maintain self-guided dietary changes. Longer interventions will clarify whether the convenience provided by meal delivery services might support longer-term dietary change, especially at times when individuals experience elevated depressive symptoms. Relatedly, follow-up periods may help to distinguish the feasibility and effectiveness of the dietary approaches themselves from the increased structure and accountability provided by the study (e.g., regular laboratory visits, frequent researcher contact, self-monitoring via EMA), which may also enhance dietary adherence and mental health outcomes.

In future applications, combined (e.g., partial meal delivery) or adaptive approaches (e.g., a shift from meal delivery to self-guided dietary change over time) may address limitations of each individual approach by creating synergistic effects and overcoming challenges in implementation such as convenience and compatibility with individual preferences. Additionally, tailoring can be used to match an intervention approach to individual needs; for example, providing the more costly meal delivery only to those with lower nutrition literacy, chronic time constraints, or extreme fatigue and avolition due to depression. Follow-up qualitative analyses of debriefing interviews from the current study that provide more insight into participants’ experiences with each intervention approach and factors contributing to their adherence will be conducted and published separately to guide these future directions.

Additional limitations and considerations

While the current study provides preliminary evidence for the feasibility and potential effectiveness of two minimally processed dietary interventions, several limitations should be considered. First, we reiterate that a primary limitation is the inability to precisely characterize participants’ nutrient-level intake or quantify the proportional contribution of provided meals to overall dietary intake. Although Diet ID allows for measurement of general self-reported dietary patterns, it does not capture exact energy intake, macronutrient composition, or adherence to study-provided foods. To improve characterization of the dietary intervention, we compiled estimated nutritional information for delivered meals using the MyNetDiary database, and list the meals provided to each participant along with available nutritional estimates in Supplemental Material. Estimated meal composition and self-reported dietary quality from Diet ID suggest increases in dietary fiber and decreases in added sugar during the intervention, consistent with what would be expected from a shift toward a more minimally processed diet⁴⁹; however, these findings should be interpreted with caution, as the MyNetDiary values could not be verified against nutrient composition information from the meal delivery company and actual consumption was not monitored. Consequently, the lack of precise intake data and uncertainty regarding adherence limit the interpretability of observed effects and the reproducibility of the intervention in future trials. Additionally, for participants in the self-implemented nutritional guidance condition, no meal-level nutritional estimates were available beyond the overall dietary pattern data captured by Diet ID, further limiting our ability to compare nutrient-level intake both pre–post intervention and across conditions.

Second, participants in this study primarily identified as female, non-Hispanic White, highly educated, and economically middle-class, which may limit applicability to broader populations. Access to minimally processed foods can be shaped by socioeconomic status, with higher-income households better able to afford and obtain fresh produce, whole grains, and other nutrient-rich options than lower-income households^71–74. Unequal division of household labor based on gender is also common, with American Time Use data demonstrating that women spend significantly more time on food preparation than men⁷⁵. For some individuals, preparing minimally processed foods or eating pre-prepared meals may inadvertently increase stress and effort associated with food preparation, especially if other household members (e.g., partner, children) have different dietary preferences and needs. Future research with larger and more diverse samples is needed to understand how these challenges may impact one’s ability to eat a minimally processed diet or utilize a meal delivery service, especially over time.

Finally, these individual-level considerations underscore the importance of broader community- and policy-level interventions aimed at improving dietary access. Emerging evidence suggests that interventions that increase the availability of and access to affordable, minimally processed foods in community settings (e.g., mobile food units) may be associated with improvements in depressive symptoms over time⁷⁶. Such approaches, alongside other structural changes to the food environment, are likely necessary for dietary and mental health interventions to be effective at scale, as they can reach individuals who lack the resources required for individualized interventions.

Conclusion

Nearly 60 percent of the daily calories consumed in the U.S. come from ultra-processed foods⁶⁶, which poses significant risks for the development and maintenance of depressive disorders^{8,10–12,14,15}. Interventions that facilitate shifts toward minimally processed foods may therefore represent promising, scalable strategies for depression prevention and treatment that also overcome challenges associated with standard care (e.g., underdiagnosis, accessibility, treatment resistance)^6–8. In particular, structured meal delivery may offer a uniquely low-burden and accessible approach by reducing cognitive, motivational, and logistical barriers that can impede dietary change, especially for individuals experiencing heightened depressive symptoms.

To our knowledge, this study provides the first test of a commercial meal delivery service as a strategy to increase the convenience of consuming a minimally processed diet and target depressive symptoms in adults with moderate to moderately severe depression. Findings lay the groundwork for future randomized controlled trials and underscore the importance of increasing access to nutrient-rich, minimally processed foods and building practical knowledge of how to incorporate them within one’s regular diet. Such efforts, at both individual and societal levels, could play a vital role in addressing depression and supporting overall mental health.

Author contributions

Conceptualization: A. Gearhardt, K. Sonneville, J. Lee; Methodology: A. Gearhardt, C. Furman, I. Worth, J. Zhang, K. Sonneville, J. Lee; Formal Analysis: C. Furman, J. Henderson; Investigation: I. Worth, J. Zhang; Resources : A. Gearhardt, C. Furman, I. Worth, J. Zhang; Data Curation: C. Furman, J. Zhang; Writing—Original Draft: A. Gearhardt, C. Furman, I. Worth, J. Zhang, J. Henderson, E. Pokowitz; Writing—Review & Editing: K. Sonneville, J. Lee; Visualization : C. Furman; Supervision: A. Gearhardt; Project Administration: A. Gearhardt, I. Worth, J. Zhang; Funding Acquisition: A. Gearhardt, K. Sonneville, J. Lee.

Funding

We would like to acknowledge the Frances and Kenneth Eisenberg Endowment Fund and the University of Michigan Eisenberg Family Depression Center for their support in providing research funding that made this project possible. REDCap at the University of Michigan is supported by the Michigan Institute for Clinical and Health Research (MICHR), which is funded by the National Center for Advancing Translational Sciences (NCATS) of the National Institutes of Health under Award Number UM1TR004404. In addition to direct funding of this project, A. Gearhardt, C. Furman, J. Zhang, I. Worth, J. Henderson, and K. Sonneville are supported by the National Institute on Drug Abuse under Award Number R01DA055027. J. Lee is supported by the Elizabeth Weiser Caswell Diabetes Institute at the University of Michigan and the National Institute of Diabetes and Digestive and Kidney Diseases under Award Numbers P30DK089503 (MNORC), P30DK020572 (MDRC), and P30DK092926 (MCDTR).

Data availability

The full protocol and measures, data for the current study aims, and analytic code can be found on the Open Science Framework project page (OSF; https://osf.io/pcufr).

Declarations

Competing interests

Principle Investigator A. Gearhardt has received speaking honoraria from academic organizations and health-related nonprofits, consulting fees from health-related nonprofits and a law firm and receives royalties from Oxford University Press. Co-I J. Lee receives grant funding from Lilly USA, LLC, participated on the Medical Advisory Board for GoodRx, and served as a consultant to Tandem Diabetes Care. These consulting roles and companies were not involved in any step of the research process (e.g., funding, study design, collection and analysis of data, writing or approval of manuscript) and did not influence the content, analysis, or conclusions of this research project or manuscript. No other conflicts of interest are declared with respect to the authorship or the publication of this article.

Ethics approval and consent to participate

This study’s procedures and materials were approved by the Institutional Review Board at the University of Michigan. All participants provided written informed consent.