Food as Medicine for Mental Health: What the SMILES Trial Means for Benefit Design

Why this matters for payers and value-based care leaders

Behavioral health spend is the fastest-growing line item in US commercial and Medicare Advantage medical-loss ratios. Nearly one in five Medicare beneficiaries carries a diagnosed depressive or anxiety disorder, and the downstream costs — higher ED utilization, worsening diabetes control, elevated cardiovascular events, medication non-adherence — compound quietly on the medical side of the P&L rather than the pharmacy side.

Payers and ACO enablers have plenty of tools for severe mental illness (SSRIs, psychotherapy, intensive outpatient programs, inpatient). What has been missing is a defensible, evidence-backed, low-cost adjunct for mild-to-moderate depression and generalized anxiety — the volume segment. Food-as-medicine is now that adjunct.

The SMILES finding, in plain language

The SMILES trial (Jacka et al., BMC Medicine, 2017) was the first randomized controlled trial to isolate the effect of a dietary intervention on moderate-to-severe depression.¹ Sixty-seven adults with an active diagnosis of major depressive disorder were randomized to either:

• Dietary support arm: seven 60-minute sessions with a clinical dietitian over 12 weeks, coaching a Mediterranean-style dietary pattern (the ModiMedDiet — vegetables, legumes, whole grains, lean proteins, olive oil, fish, minimal ultra-processed foods).
• Social support arm: seven matched sessions of befriending / social contact, no dietary content.

After 12 weeks, the dietary support arm showed a clinically significant improvement on the Montgomery-Åsberg Depression Rating Scale (MADRS), with 32.3% of diet participants meeting remission criteria versus 8.0% in the social-support arm. The number needed to treat was approximately 4 — comparable to or better than many first-line antidepressant comparisons.¹

What SMILES does not claim

SMILES was not a first-line-therapy replacement trial. It was adjunctive — participants remained on their existing treatment regimens, including antidepressants and psychotherapy, throughout the study. The intervention arm added dietary support on top of usual care. The right framing is therefore: food-as-medicine adds to, rather than substitutes for, pharmacological and psychotherapeutic treatment.

SMILES also was not a large trial. Sixty-seven participants is a feasibility-scale cohort. What it did do is establish the mechanistic plausibility and the effect-size ballpark that has since been replicated and extended by other groups.

The replication signal

Since SMILES, multiple trials have tested variants of the hypothesis. Results are mixed but directional:

• HELFIMED (Parletta et al., Nutr Neurosci 2019) — Australian RCT of 152 adults with depression randomized to Mediterranean diet + fish oil vs social groups; dietary arm showed sustained improvement in depression scores at 3 and 6 months.
• MoodFOOD (Bot et al., JAMA 2019)² — large multicountry prevention trial in overweight adults with elevated depressive symptoms. Behavioral activation helped; food supplementation in isolation did not. Reinforces that the active ingredient is dietary pattern and behavior change, not specific supplements.
• SUN Project cohort — observational Spanish cohort consistently showing inverse association between Mediterranean diet adherence and incident depression.
• Nurses' Health Study II — similar inverse-association signal in a large US cohort.

Taken together: dietary pattern is a real adjunctive lever for depression, with strongest evidence for Mediterranean and MIND dietary patterns, and with effect sizes large enough to move clinical scales at the population level.

What this means for benefit design

For MA plans, ACO enablers, and Medicaid plans operating under HRSN waivers, the benefit-design implications are:

1. Behavioral-health food-as-medicine meets the SSBCI evidence bar. CMS CY2025 rules require documented evidence that a Special Supplemental Benefit for the Chronically Ill has a reasonable expectation of improving or maintaining health or overall function. SMILES plus the replication set clears that bar for members with diagnosed mild-to-moderate major depressive disorder and generalized anxiety.
2. It compounds with medical-side outcomes in chronic-disease patients. Depression is the single strongest predictor of medication non-adherence in diabetes and hypertension. Improving mood improves adherence, and improving adherence lowers medical cost. Food-based behavioral-health support has the rare property of being beneficial on both the pharmacy-adjacent and the medical sides.
3. It pairs cleanly with validated PROM capture. PHQ-9 and GAD-7 on a 14-day cadence produce the outcome signal that value-based contracts reward, whether through supplemental-benefit accounting, Star measures, or outcomes-linked arrangements with ACO enablers.
4. It is scope-bounded. Food-as-medicine for behavioral health is not psychotherapy, medication management, or crisis intervention. It is adjunct. Contract language and member-facing disclaimers need to say so. Any vendor making the opposite claim is uninsurable.

What RxPulse + RxDiet do with this evidence base

We translate the SMILES / MoodFOOD / HELFIMED literature into a delivered adjunct, not a trial:

• RxDiet menu library includes a Mood Support tag set — high-omega-3, high-folate, anti-inflammatory, gut-supportive, fermented, magnesium-rich recipes. Members on a BH pack receive meal plans that default to these tags, without requiring them to read the literature.
• Allie (our AI companion) delivers behavioral activation coaching — scheduling pleasant activities, tracking mood, sleep hygiene, 5-4-3-2-1 grounding, meal-based mood anchors. Evidence-backed session structures informed by the SMILES dietitian protocol, delivered at scale via voice and chat.
• Validated PROMs (PHQ-9, GAD-7) administered on a 14-day cadence — enough frequency to detect response without creating survey fatigue. Every submission is stored with full audit trail, and our FHIR MeasureReport export is ready for any value-based contract that requires it.
• Item-9 PHQ-9 escalation is non-negotiable. Any suicidality signal triggers a crisis-resource modal (988 in the US) and a clinical-event record; Allie does not attempt crisis counseling under any circumstance. Our entire BH architecture is built around this scope boundary.

The next trials to watch

Two research directions are active and worth monitoring for any payer or value-based care leader shaping future behavioral-health benefits:

• Psychobiotics and the gut-brain axis — fermented-food and probiotic strain-specific trials in mood disorders. Evidence thinner than SMILES but accumulating.
• Conversational-AI-delivered behavioral activation— early narrative reviews (Im & Woo, JMIR Mental Health 2025)³ suggest that chatbot-delivered CBT can be clinically meaningful for mild-to-moderate populations. RCTs specifically on voice-delivered vs text-delivered interventions are underway but not yet published.

References

1. Jacka FN, O'Neil A, Opie R, et al. A randomised controlled trial of dietary improvement for adults with major depression (the SMILES trial). BMC Medicine. 2017;15:23. doi:10.1186/s12916-017-0791-y. PMID: 28137247.
2. Bot M, Brouwer IA, Roca M, et al. Effect of Multinutrient Supplementation and Food-Related Behavioral Activation Therapy on Prevention of Major Depressive Disorder Among Overweight or Obese Adults With Subsyndromal Depressive Symptoms: The MooDFOOD Randomized Clinical Trial. JAMA. 2019;321(9):858-868. doi:10.1001/jama.2019.0556.
3. Im SS, Woo JH. Digital Mental Health Interventions Using CBT Chatbots for Depression and Anxiety: A Narrative Review. JMIR Mental Health. 2025. PMC12669916.

Additional supporting literature archived in our internal evidence library: Downer et al. (BMJ 2020) — framework paper on food-is-medicine; HELFIMED; SUN Project cohort analyses; Nurses' Health Study II depression-diet analyses. Full evidence-library index available on request under our editorial COI disclosure.