Full mechanism, in text

A trial of eating differently

1 · design

The trial randomised 67 people with moderate-to-severe depression to coached modified Mediterranean diet or social-support control for 12 weeks, on top of their existing treatment.Jacka 2017 BMC Medicine SMILES Trial — the strongest for dietary intervention in depression on this map:

Study design:

· N = 67 clinically diagnosed depression (DSM-IV-TR major depressive episode)
· Entry required MADRS ≥ 18 plus a Dietary Screening Tool score confirming poor diet quality; mean baseline MADRS 25.4
· 12 weeks, 1:1 randomisation
· Real psychiatric patients (most already on / in treatment)

Key features:

· Everyone continued existing treatment (medication / psychotherapy) — SMILES is adjunctive, not a replacement
· Intervention + control both had 7 sessions (first four weekly, the remaining three fortnightly), frequency and duration matched → controls for attention placebo
· Primary endpoint: MADRS (Montgomery-Åsberg Depression Rating Scale) 12-week change

Intervention group (N=33):

ModiMedDiet (modified Mediterranean), individually coached by RD (registered dietitian) in 7 sessions:

· Daily: 6 servings vegetables + 3 fruit + 5-8 whole grains + 3-4 legumes/week + 1 serving nuts
· Weekly: fish ≥ 2× + red meat 3-4× (lean only) + poultry 2-3× + up to 6 eggs
· Oil: mainly olive oil (3 tbsp/day)
· Dairy: low-fat 2-3 servings/day
· Limit: (sweets / baked / fried / processed meat / SSB / drippings) → < 3 servings/week

Control group (Social Support, N=34):

· Befriending protocol — neutral social activities (cards / walks / TV chat / weather)
· No dietary or psychological content
· Same visit frequency + duration + attention as intervention

Why this control design is gold standard:

· Depression itself has high placebo + attention effect (40% response rate common)
· Befriending equally gives attention + ritual + expectation — intervention group's true benefit = the diet itself
· N=67 small but rigorous: subsequently HELFIMED + other replications (Parletta 2019)

2 · MADRS trajectory

MADRS fell 11.3 points in the intervention group and 4.2 in the control group, a 7.1-point gap, measured only before the programme and at week 12.Primary endpoint — MADRS 12-week change:

MADRS = Montgomery-Åsberg Depression Rating Scale (0-60, higher = more severe)

· 0-6: normal
· 7-19: mild ← the SMILES entry threshold sits in THIS band: MADRS ≥ 18
· 20-34: moderate ← where the enrolled cohort actually averaged (baseline 25.4)
· ≥ 35: severe

SMILES results:

· Baseline: intervention MADRS 26.1, control 24.7
· Week 12: intervention 14.8, control 20.5
· MADRS change: intervention −11.3 points · control −4.2 points
· Between-group difference 7.1 points (95% 3.9-10.4, t(60.7) = 4.38, p < 0.001, Cohen's d = 1.16)
· ⚠️ The paper publishes a confidence interval for the between-group difference only — not for either within-arm change

Cohen's d interpretation:

· d = 0.2 small, 0.5 medium, 0.8 large
· d = 1.16 = large effect
· ⚠️ Do not size this against antidepressants: SMILES is a small trial with 56 completers in which participants knew their group but the assessors who scored them did not, while the reference work on drugs (Cipriani 2018 network ) reports for response, not Cohen's d. The two numbers are not on the same ruler, so a claim that it is bigger than does not hold

Why the control group also dropped 4.2 points:

· Natural course + attention + expectation + existing treatment (most on SSRI / )
· Depression's regression to the mean: enrolled at trough, any time-point is more likely to revert
· This is why compare with 0 isn't enough; must compare with matched attention control

⚠️ There are only two measurement points: SMILES assessed participants before the programme and again at week 12, with nothing in between. So this line is two points joined, not a trajectory — week 6 was never measured, so the paper contains no already significant at week 6 result

HADS-A (anxiety) also improved:

· Intervention 12.1 → 8.4, control 11.2 → 9.5
· Between-group difference 2.0 points (95% CI 0.2-3.9, p = 0.033, Cohen's d = 0.59)
· Effective for depression + anxiety comorbidity too, but the anxiety effect is far smaller than the depression one (d 0.59 vs 1.16)
· ⚠️ The instrument was HADS-A, not HAM-A — the Hamilton Anxiety scale was never administered in this trial

3 · remission + NNT

At week 12, 32.3% of the intervention group remitted versus 8.0% of controls, about 1 extra remission per 4 people who changed diet, as an adjunct to existing treatment.Remission defined as MADRS ≤ 9 at week 12:

· Intervention: 32.3% (10/31 completers)
· Control: 8.0% (2/25 completers)
· p = 0.028, = 5.5

(Number Needed to Treat) = 4.1

· Meaning: out of every 4 moderate-to-severe depression patients who change diet via SMILES, an average of 1 extra patient remits (vs continuing standard treatment + attention control)
· ⚠️ Do not rank this NNT against antidepressants: the reference work on drugs (Cipriani 2018 network ) reports odds ratios for response, not NNTs, and SMILES is a small trial with 56 completers in which participants knew their group but the assessors who scored them did not. Calling it lower than drugs puts two different rulers side by side

Clinical Global Impression (CGI-I):

· Intervention mean 2.1 (much improved) vs control 3.0 (minimally improved), t(50) = −2.58, p = 0.013
· ⚠️ SMILES ran no analysis of response defined as a ≥ 50% MADRS reduction, so the paper reports no response rate

Dietary adherence (ModiMedDiet score, out of 120 — not 12):

· Intervention 36.2 → 55.1
· Control 47.3 → 45.4 (⚠️ the control arm started higher, so these are not two lines from the same origin)
· Between-group difference 20.7 points (p < 0.001)
· The more the diet changed, the more depression fell: each 10% gain in adherence went with about 2.2 more points off MADRS

Economics:

· Cost: the paper modelled the diets of 20 participants — AU$138/week at baseline vs AU$112/week for the recommended diet, about AU$26/week cheaper
· ⚠️ That is a modelled weekly grocery cost, not a measured three-month spend, and not a promise of how much you save per month
· The change is a substitution ( → fruit/veg + fish), not adding expensive items — that part the paper does support

Replications (post-SMILES):

· Parletta 2019 HELFIMED: 152 recruited, 95 completing at 3 months and 85 at 6; participants had self-reported depression, not a clinical diagnosis of severe depression. Mediterranean diet plus fish oil vs social groups → larger fall in depression scores (p = 0.03), still present at 6 months. ⚠️ Fish oil is part of the intervention, so this is not a clean diet-only replication
· Francis 2019: 101 young adults of both sexes with depressive symptoms, 3 weeks of a Mediterranean-style diet vs habitual diet → CESD-R (p = 0.007) and DASS-21 depression (p = 0.002) both fell
· Bayes 2022 AMMEND: 72 Australian men aged 18-25 completing, 12-week Mediterranean diet vs befriending control, BDI-II between-group difference 14.4 points (p < 0.001). ⚠️ It is a single open-label single-site , not a review of 20+ RCTs

Key: SMILES is adjunctive, not a replacement:

· People on antidepressants + psychotherapy continue, add diet → a third leg
· Do NOT stop medication just because diet improved — this is the most dangerous suicide-risk pitfall in SMILES replications
· Severe depression (PHQ-9 ≥ 20) / suicidal ideation: psychiatry / psychology first, nutrition adjunctive

4 · mechanisms

Five hypothesised pathways may link diet to depression: gut microbiome, low-grade inflammation, one-carbon metabolism, blood glucose stability and behavioural activation.SMILES — 5 mechanistic hypotheses:

① Gut microbiome — gut-brain axis:

· Mediterranean's high fibre + polyphenols → (short-chain fatty acids) ↑
· SCFA → gut mucosa stability + vagal nerve + suppresses inflammatory cytokines
· Depressed people have lower gut microbiota richness and diversity (Kelly 2016), and Coprococcus and Dialister are specifically depleted — still true after correcting for antidepressant use (Valles-Colomer 2019)
· Faecal microbiota transplant (FMT): transferring microbiota FROM depressed patients INTO microbiota-depleted rats induced anhedonia, anxiety-like behaviour and altered tryptophan metabolism in the recipients (Kelly 2016)
· ⚠️ The direction is transfer, not treatment — it is precisely because the phenotype travels with the microbiota that this suggests causality. Saying FMT improved depressed rats gets both the direction and the result backwards

② Low-grade inflammation ↓:

· + high sugar + ω-6 overload → systemic low-grade chronic inflammation
· Depressed people have higher mean / / IL-1 (Howren 2009 ) — but the effects are small: CRP d = 0.15, IL-6 d = 0.25. ⚠️ is not in this meta (that is Dowlati 2010, a different paper)
· Inflammation-cytokine model of depression — leading hypothesis
· Mediterranean → anti-inflammatory (olive oil + fish + polyphenols) → inflammatory markers drop → depression improves pathway

③ One-carbon metabolism + neurotransmitter synthesis:

· (serotonin) synthesis needs tryptophan + B6
· DA (dopamine) + NE synthesis need tyrosine + Fe + Cu + vitamin C
· universal methyl donor needs B12 + folate + betaine
· B-vitamin insufficiency + depression strongly correlated (especially elderly)

④ Blood glucose stability:

· Postprandial glucose spike + insulin drop: fluctuating brain energy supply
· GLUT3 (brain + neurons) is insulin-independent: but glucose < 60 affects cognition / mood
· Mediterranean = low + high fibre → stable glucose

⑤ Behavioural activation + self-efficacy (psychology):

· I can change my body is core to depression treatment ( behavioural activation module)
· Learning to cook + shopping + dining social → behaviour + social + sense of accomplishment
· SMILES RD 1:1 coaching: not here's a list but learning together → psychological effect also counts

Practical 12-week protocol:

Weeks 1-2 add / substitute:

· Vegetables 4-6 servings/day + fruit 2-3 + whole grains ≥ 3 + fish 2-3×/week + nuts 30 g/day

Weeks 3-4 switch oils:

· Olive oil as main oil · reduce ω-6 (soy / sunflower / corn) · butter / lard < 1 tbsp/day

Weeks 5-12 protein source structure:

· Legumes 3×/week · eggs moderate · poultry + fish replace most red meat · processed meat minimal

Quit:

· UPF (5+ additives / hydrogenated oils / ) · SSB · fast food / fried

Don't:

· Don't believe diet alone cures depression (post-SMILES online exaggeration; the paper says adjunct)
· Don't expect full remission at 12 weeks (32% > 0 but the rest still need to continue treatment)
· Don't keto / extreme-low-carb for depression (single signal, long-term adherence very poor)
· Don't stop / antidepressant because diet changed — must discuss with psychiatry

Suicide risk:

· If suicidal ideation → psychiatry / psychology is the mainline, diet adjunctive
· Depression hotlines (China 12356 / US 988)
· Any clear today I can't go on → don't go it alone, seek help

Full mechanism, in text

1 · design

2 · MADRS trajectory

3 · remission + NNT

4 · mechanisms

SMILES Trial · Jacka 2017 · Mediterranean diet 32% remission

Full mechanism, in text

1 · design

2 · MADRS trajectory

3 · remission + NNT

4 · mechanisms

SMILES Trial · Jacka 2017 · Mediterranean diet 32% remission