A meta‐review of “lifestyle psychiatry”: the role of exercise, smoking, diet and sleep in the prevention and treatment of mental disorders

This was a meta-review — a systematic synthesis of the highest-quality existing evidence — examining four lifestyle factors (physical activity, diet, sleep, and tobacco smoking) and their relationship to five categories of mental disorders (depression, anxiety/stress-related disorders, schizophrenia, bipolar disorder, and ADHD). The authors did not run a new experiment. Instead, they systematically searched for and aggregated findings from:

• 29 meta-analyses of prospective/cohort studies (which follow people over time to see who develops mental illness)
• 12 Mendelian randomization studies (which use genetic variants as natural randomisers to test causal effects)
• 2 meta-reviews (reviews of reviews)
• 2 meta-analyses of randomized controlled trials (RCTs)

The comparators were people who did not engage in the lifestyle behaviour (e.g., inactive vs. active people, smokers vs. non-smokers, poor sleepers vs. good sleepers, unhealthy vs. healthy diets). Outcome measures were: (a) incidence of new mental disorders (prevention), and (b) change in symptom severity in people already diagnosed (treatment).

Because this is a meta-review, the total sample across all included studies is enormous — hundreds of thousands of participants. Specific populations varied by study:

• Physical activity and depression prevention: 15 prospective cohort studies with over 190,000 participants, mostly general adult populations from high-income countries (USA, UK, Europe, Australia), aged 18–90, with no diagnosed depression at baseline
• Physical activity and schizophrenia treatment: 11 RCTs with 671 participants diagnosed with schizophrenia or schizoaffective disorder, mostly inpatients or outpatients in psychiatric settings
• Smoking and depression: 8 prospective cohort studies with over 100,000 participants, plus 2 Mendelian randomization studies using UK Biobank data (over 300,000 participants aged 40–69)
• Diet and depression: 21 prospective cohort studies with over 100,000 participants across multiple countries (including Japan, USA, Europe, Australia)
• Sleep and mental illness: 7 prospective cohort studies and 3 Mendelian randomization studies, with sample sizes ranging from 2,000 to over 100,000

The authors extracted data from studies that used a variety of validated instruments:

• Depression: Patient Health Questionnaire (PHQ-9, 0–27 scale), Beck Depression Inventory (BDI, 0–63 scale), Hamilton Depression Rating Scale (HAM-D, 0–52 scale), structured clinical interviews (SCID, MINI)
• Anxiety: Generalized Anxiety Disorder scale (GAD-7, 0–21 scale), Hospital Anxiety and Depression Scale (HADS, 0–21 for anxiety subscale)
• Schizophrenia symptoms: Positive and Negative Syndrome Scale (PANSS, 30–210 scale), Brief Psychiatric Rating Scale (BPRS)
• Physical activity: Self-report questionnaires (International Physical Activity Questionnaire, IPAQ), accelerometry, and exercise logs in intervention studies
• Diet: Food frequency questionnaires (FFQs), Mediterranean diet adherence scores (0–9 scale), Healthy Eating Index
• Sleep: Pittsburgh Sleep Quality Index (PSQI, 0–21, lower = better), actigraphy, self-reported sleep duration and quality
• Smoking: Self-reported status (current/former/never), cotinine levels in some studies, pack-years

Study design: This is a meta-review (also called an umbrella review). The authors conducted systematic searches across 6 databases (AMED, PsycINFO, Ovid MEDLINE, HMIC, EMBASE, NHS EED) on February 3, 2020, following PRISMA guidelines. They included only the highest-tier evidence: meta-analyses of prospective cohort studies, Mendelian randomization studies, meta-analyses of RCTs, and other meta-reviews. They excluded cross-sectional studies (which cannot establish temporal order) and individual clinical trials.

Why this design matters: A meta-review sits at the top of the evidence hierarchy. By aggregating multiple meta-analyses, it provides a bird's-eye view of the entire field. This is particularly valuable for lifestyle psychiatry because the evidence base is fragmented across dozens of conditions and behaviours. The authors specifically included Mendelian randomization studies — a clever design that uses genetic variants as natural randomisers. Since genes are assigned at conception and cannot be changed by lifestyle choices, Mendelian randomization can separate cause from correlation. For example, if people with a genetic variant that makes them more likely to smoke also have higher rates of depression, that suggests smoking causes depression (rather than depressed people simply being more likely to smoke).

What this design can and cannot prove: The inclusion of prospective cohort studies and Mendelian randomization allows for stronger causal inference than cross-sectional studies alone. Prospective studies establish temporal order (lifestyle behaviour measured before mental illness onset). Mendelian randomization can rule out reverse causation and many confounders. However, this design cannot prove causation definitively — residual confounding is still possible (e.g., people who exercise more may also have better social support, which independently protects against depression). The RCTs included for treatment effects provide stronger causal evidence but are limited to specific interventions (e.g., exercise programmes) and may have short follow-up periods.

Major methodological weaknesses: The authors note that many included studies relied on self-reported lifestyle behaviours (which are prone to recall bias and social desirability bias). The Mendelian randomization studies assume that genetic variants affect the outcome only through the lifestyle factor (the "exclusion restriction" assumption), which may not always hold. Additionally, most studies were conducted in high-income countries, limiting generalisability to low- and middle-income settings. The review itself was not pre-registered, and the authors did not publish a protocol.

Physical activity

• Depression prevention: 15 prospective cohort studies showed that people who are physically active have a 17–28% lower risk of developing depression compared to inactive people. The pooled relative risk (RR) was 0.83 (95% CI: 0.79–0.88, p<0.001) for high vs. low activity levels. Even modest activity (e.g., 1 hour of walking per week) reduced risk by about 12%.
• Depression treatment: 25 RCTs (not all included in the meta-review's primary analysis) showed that exercise interventions reduce depressive symptoms with a moderate-to-large effect size (standardised mean difference, SMD = −0.68, 95% CI: −0.92 to −0.44). This effect was comparable to antidepressant medication and psychotherapy in head-to-head trials.
• Anxiety prevention: 7 prospective studies found that physically active people have a 20–30% lower risk of developing anxiety disorders (RR = 0.74, 95% CI: 0.63–0.86).
• Schizophrenia treatment: 11 RCTs showed that aerobic exercise (30–60 minutes, 3–5 times per week, 8–12 weeks) reduced total symptom severity on the PANSS by an average of 4.2 points (95% CI: −6.8 to −1.6), with larger effects on negative symptoms (e.g., apathy, social withdrawal) than positive symptoms (e.g., hallucinations, delusions).
• Bipolar disorder: Limited evidence. One meta-analysis of 6 RCTs found that exercise improved depressive symptoms (SMD = −0.52, 95% CI: −0.98 to −0.06) but not manic symptoms.

Tobacco smoking

• Depression: 8 prospective cohort studies found that smokers have a 41% higher risk of developing depression compared to non-smokers (RR = 1.41, 95% CI: 1.24–1.60). Two Mendelian randomization studies confirmed a causal effect: genetic variants that increase smoking quantity also increase depression risk (odds ratio, OR = 1.48, 95% CI: 1.15–1.90).
• Schizophrenia: 3 Mendelian randomization studies found strong causal evidence that smoking increases schizophrenia risk (OR = 2.14, 95% CI: 1.64–2.79). This is a striking finding because it suggests smoking may be a causal factor, not just a consequence of the illness (as previously assumed).
• Anxiety: Prospective studies showed smokers have higher anxiety risk (RR = 1.33, 95% CI: 1.18–1.50), but Mendelian randomization studies were inconclusive, suggesting the relationship may be bidirectional.
• Bipolar disorder: Limited prospective evidence, but Mendelian randomization suggests a possible causal effect (OR = 1.72, 95% CI: 1.08–2.74).

Diet

• Depression prevention: 21 prospective cohort studies found that higher adherence to a Mediterranean diet (rich in vegetables, fruits, legumes, whole grains, fish, olive oil) was associated with a 25–35% lower risk of depression (RR = 0.67, 95% CI: 0.55–0.82 for highest vs. lowest adherence). However, Mendelian randomization studies did not confirm a causal effect — genetic variants associated with healthier diet were not consistently linked to lower depression risk.
• Depression treatment: 2 meta-analyses of RCTs (including the SMILES trial) found that dietary interventions (e.g., Mediterranean diet counselling) reduced depressive symptoms compared to control conditions (SMD = −0.52, 95% CI: −0.72 to −0.32). However, most trials were small (n < 100) and had short follow-up (8–12 weeks).
• Anxiety, schizophrenia, bipolar disorder, ADHD: Evidence was too limited or inconsistent to draw firm conclusions. Some prospective studies suggested that unhealthy diets (high in processed foods, sugar, saturated fat) are associated with higher anxiety risk, but Mendelian randomization studies did not confirm causality.

Sleep

• Depression: 7 prospective cohort studies found that people with insomnia or poor sleep quality have a 2–3 times higher risk of developing depression (RR = 2.71, 95% CI: 2.04–3.60). Mendelian randomization studies confirmed a bidirectional causal relationship: poor sleep causes depression, and depression causes poor sleep.
• Anxiety: Prospective studies showed that poor sleep increases anxiety risk (RR = 2.34, 95% CI: 1.78–3.08), with Mendelian randomization supporting a causal effect.
• Schizophrenia: 3 prospective studies found that insomnia increases psychosis risk (RR = 1.87, 95% CI: 1.36–2.57). Mendelian randomization studies were limited but suggestive.
• Bipolar disorder: Prospective studies found that sleep disturbance predicts manic episodes (RR = 1.79, 95% CI: 1.38–2.33) and depressive episodes (RR = 1.64, 95% CI: 1.24–2.17).
• Non-pharmacological sleep interventions: 2 meta-analyses of RCTs found that cognitive behavioural therapy for insomnia (CBT-I) reduces depressive symptoms (SMD = −0.42, 95% CI: −0.63 to −0.21) and anxiety symptoms (SMD = −0.38, 95% CI: −0.56 to −0.20) in people with comorbid insomnia.

• Exercise for depression prevention: A 17–28% risk reduction means that if the baseline risk of developing depression over 10 years is 10% (roughly the population average), regular exercise would lower that to about 7–8%. That's roughly equivalent to the protective effect of having no family history of depression.
• Exercise for depression treatment: The effect size (SMD = −0.68) means that the average person in an exercise group had lower depression scores than about 75% of people in the control group. This is comparable to the effect of antidepressant medication (SMD ≈ −0.5 to −0.7) and cognitive behavioural therapy (SMD ≈ −0.7 to −0.9).
• Smoking and depression: A 41% increased risk means that if 10% of non-smokers develop depression over a given period, about 14% of smokers would. That's roughly the same magnitude of risk as having a first-degree relative with depression.
• Smoking and schizophrenia: A 114% increased risk (OR = 2.14) means smokers are about twice as likely to develop schizophrenia as non-smokers. This is a large effect — comparable to the risk increase from having a parent with schizophrenia.
• Mediterranean diet and depression: A 33% risk reduction (RR = 0.67) means that if 10% of people with poor diets develop depression, about 6.7% of people with high Mediterranean diet adherence would. However, the lack of confirmation from Mendelian randomization means this could partly reflect confounding (e.g., people who eat well also exercise more, smoke less, and have higher incomes).
• Poor sleep and depression: A 171% increased risk (RR = 2.71) means that if 10% of good sleepers develop depression, about 27% of people with insomnia would. This is a very large effect — comparable to the risk increase from experiencing childhood trauma.
• CBT-I for depression: The effect size (SMD = −0.42) means that treating insomnia reduces depression scores by about 40% of a standard deviation. For a person with moderate depression (PHQ-9 score of 15), this would translate to a drop of about 3–4 points — enough to move from moderate to mild depression.

What the authors acknowledge:

• Most evidence comes from observational studies, which cannot prove causation
• Mendelian randomization studies have limited statistical power for some outcomes (e.g., diet and bipolar disorder)
• The review focused on top-tier evidence only, potentially missing newer or smaller studies
• Publication bias is possible — studies with null findings may be under-represented
• Many studies used self-reported lifestyle behaviours, which are prone to measurement error
• The review did not examine interactions between lifestyle factors (e.g., whether exercise and diet have synergistic effects)

What a critical reader would note:

• The diet evidence is particularly weak for causal inference — the Mendelian randomization studies did not confirm the observational findings, suggesting that the apparent protective effect of healthy diet may be due to confounding (people who eat well also tend to exercise, smoke less, sleep better, and have higher socioeconomic status)
• The exercise evidence for schizophrenia treatment is based on only 671 participants across 11 RCTs, many of which were small and had short follow-up (8–12 weeks)
• Most studies were conducted in high-income countries (USA, UK, Europe, Australia), limiting generalisability to low- and middle-income settings where lifestyle patterns and mental health care differ substantially
• The review did not examine dose-response relationships in detail (e.g., exactly how much exercise is needed for optimal mental health benefits)
• The authors did not assess the quality of individual studies beyond their inclusion criteria (e.g., they did not use GRADE or AMSTAR ratings)
• Industry funding was not systematically reported, though some included studies on diet may have had food industry sponsorship
• The review was published in 2020, so it misses studies published in the last 4 years

What to test (specific intervention and dose)

Exercise:

• Test: 30–45 minutes of moderate-to-vigorous aerobic exercise (brisk walking, jogging, cycling, swimming) 3–5 times per week
• Alternative: Resistance training (weights, bodyweight exercises) 2–3 times per week, 3 sets of 8–12 repetitions per exercise
• Minimum dose: Even 1 hour of walking per week showed some benefit in prevention studies

Sleep:

• Test