What we can piece together

After twelve parallel reviews of the literature, the factors sort cleanly by evidence certainty. The best-proven causes are mostly social, developmental and lifestyle — not exotic chemicals — and almost everything routes through the inflammation–stress–oxidation funnel.

The evidence tiers

Causal core — act here first

Supported by genetics and/or trials: childhood adversity, loneliness, smoking, physical inactivity (exercise protects, effect comparable to antidepressants), insomnia/circadian disruption, neuroinflammation (IL-6), metabolic dysfunction.

Robust contributors

Air pollution (PM2.5), ultra-processed food, alcohol, poverty/food insecurity, discrimination, job strain, HPA/chronic stress, omega-3 insufficiency, and (protective) a Mediterranean/whole-food diet.

Moderate / context-dependent

Cadmium, lead, pesticides, endocrine disruptors, mould (partly psychosocial), gut dysbiosis, vitamin-D deficiency, sugar load, adolescent cannabis, sedentary/TV, urbanicity/greenspace.

Emerging frontier

Microplastics/nanoplastics (found bioaccumulating in human brain tissue in 2024; animal models show depressive-like behaviour — human causation unresolved), aluminium, blood–brain-barrier permeability, metabolic psychiatry. Explore the frontier →

Contested / overstated

Serotonin "imbalance", PFAS, mercury-from-fish, social-media-as-primary-cause. Myth corrections →

By the numbers

~332M

people with depression (GBD 2021)

>50%

of cases begin by age ~25

~2:1

female : male prevalence

~US$1T

/yr lost productivity (with anxiety)

~$1→$4

return on scaling up treatment (WHO)

~30% / 67%

remission: first step / after 4 steps

Absolute cases rose ~59% (1990–2019) mainly via population growth/ageing; age-standardised rates are roughly flat — except youth, which are genuinely rising. COVID-19 added ~53M cases in 2020. Treatment reaches ~33% in high-income vs ~8% in low-income countries. Sources & detail →

The integrated picture

The realistic future of discovery is interaction-mining — finding combinations like low-fibre diet × high cadmium × poor sleep × high genetic risk across pooled datasets — not one exposure at a time. Ultra-processed food is a good example of bundling: it delivers additives, microplastics, aluminium, low fibre and high glycaemic load at once.

The highest-leverage bundle for most people: don't smoke, move daily, protect sleep, eat a whole-food fibre-rich diet, build social connection, and treat trauma and chronic stress — with targeted toxicant reduction and inflammation-guided treatment as secondary, precision layers.

Full write-up: download the comprehensive synthesis (PDF/Word) →