Physical Activity and Depression and Anxiety Disorders: A Systematic Review of Reviews and Assessment of Causality

The study addresses whether physical activity reduces the incidence of depression and anxiety across the life course and whether such associations are likely causal. Mental disorders, particularly depression and anxiety, are major contributors to global disability and economic burden. Prior work shows PA reduces symptoms and has been proposed as adjunct therapy, and prospective studies suggest PA may prevent onset. Existing reviews often focus on symptom reduction or specific age groups. This umbrella review aims to comprehensively summarize measures of association between PA and incident depression/anxiety across all ages, and evaluate causality using Bradford Hill criteria, to inform prevention strategies and health economic modeling (e.g., active transport health benefits).

Prior systematic reviews and meta-analyses have examined PA and mental health, often focusing on symptom reduction or specific populations. Dale et al. (2019) summarized PA effects on depression, anxiety, and self-esteem in youth but did not report strength of association for incidence and examined symptom reduction. Biddle et al. (2019) assessed causality for children/adolescents, focusing largely on symptom changes and not incident cases, finding partial support for a causal link with depression and unclear evidence for anxiety due to limited primary studies. Additional meta-reviews (Firth et al., 2020; Hoare et al., 2021) support PA in prevention and treatment of mental disorders and suggest PA can prevent depression. This umbrella review uniquely focuses on incident depression and anxiety across all ages, providing pooled measures of association and a causality assessment.

Design: Two-step protocol: (1) systematic review of reviews to establish epidemiologic evidence on the association between PA and incident depression/anxiety; (2) assessment of likelihood of causality using Bradford Hill criteria and World Cancer Research Fund grading (convincing/probable/possible/insufficient). Protocol prepared prior to conducting the review following PRISMA-P 2015 and systematic review guidelines; not registered. Databases and search: Embase and PubMed searched for reviews published from January 1, 2000 to March 19, 2020; updated to October 15, 2022. Search covered PA/active transport and mental health outcomes (depression, anxiety). MeSH terms informed search; iterative development with librarian input. Reference lists of included reviews were screened. Inclusion criteria: Systematic reviews with or without meta-analysis; exposure: physical activity; outcomes: incident depression/anxiety; participants representative of general populations without depression/anxiety at baseline (all ages or specific age groups representative of those ages); English language; peer-reviewed; reports providing risk estimates (RR, HR, OR) with CIs/SEs or data to calculate these. Study selection: Records imported to EndNote x9; duplicates removed. Titles/abstracts screened by MNW, checked by HM and JLV; full texts screened by MNW and HM; final inclusion decided by all authors, with reasons for exclusion documented. Updated search records imported to Covidence; screening by MNW and checked with JLV. Data extraction and synthesis: Extraction items agreed by four reviewers; data extracted by MNW and cross-checked by HM and JLV. Narrative synthesis adhering to PRIOR (Preferred Reporting Items for Overviews of Reviews). Quality assessment: AMSTAR rating scale by two reviewers (MNW, FH) independently. Publication bias and heterogeneity assessments as reported in included reviews (e.g., Begg/Egger, Duval and Tweedie trim-and-fill; Q and I²). Causality assessment using Bradford Hill criteria and WCRF grading. PRISMA results (initial search): 770 records (Embase=418; PubMed=352); 181 duplicates removed; 589 screened; 578 excluded; 11 full texts assessed; 7 excluded; 4 reviews included for data extraction. Updated search (2022): 669 records; 479 after duplicates; 454 excluded at title/abstract; 25 full texts; 5 reviews included (4 depression, 1 anxiety).

- Included reviews and quality: Four reviews included in the initial overview—two high quality (Schuch et al. 2018 on incident depression; Schuch et al. 2019 on incident anxiety) and two low quality (Mammen & Faulkner 2013; McDowell et al. 2019). Primary studies were prospective cohorts, largely from high-income countries, with follow-up ≥1 year. - Incident depression: Schuch et al. (2018) meta-analysis of 49 primary studies (average 7.4 years follow-up) found high PA associated with lower incidence of depression versus low PA: adjusted RR=0.83 (95% CI=0.76, 0.90; I²=0.00); adjusted OR=0.83 (95% CI=0.79, 0.88; I²=0.00). Protective effects observed across youths, adults, and older adults, and across regions. After correcting for publication bias, associations remained significant (adjusted OR=0.85, 95% CI=0.81, 0.89; adjusted RR=0.86, 95% CI=0.78, 0.96). Mammen & Faulkner (2013) reported significant inverse associations in 25/30 studies, with examples of dose-related reductions (e.g., 10–29 min/day RR=0.90; 60–90 min/day RR=0.84; >90 min/day RR=0.80). - Incident anxiety: Schuch et al. (2019) meta-analysis (14 cohorts) showed high self-reported PA associated with lower odds of incident anxiety vs low PA: adjusted OR=0.74 (95% CI=0.62, 0.88; I²=23.96); after adjusting for publication bias, AOR=0.86 (95% CI=0.69, 0.99). McDowell et al. (2019) reported reduced odds across outcomes: self-reported anxiety symptoms mean AOR=0.87 (95% CI=0.77, 0.99; I²=48.67%); diagnosis of any anxiety disorder mean OR=0.66 (95% CI=0.53, 0.82; I²=62.26%); generalized anxiety disorder mean OR=0.54 (95% CI=0.32, 0.92; I²=0%). - Heterogeneity and publication bias: Heterogeneity ranged from low to moderate/high depending on outcome; publication bias detected in several analyses, with trim-and-fill adjustments preserving significance. - Dose-response: Evidence of dose-response for depression reported in several primary studies (e.g., increasing daily PA linked to progressively lower risk). For anxiety, McDowell & Dishman reported 11 studies showing lower odds with increased PA amounts; Schuch & Stubbs did not assess dose-response. - Updated search (to Oct 15, 2022): Additional reviews largely corroborated associations and provided dose-response insights. Pearce et al. (2022) meta-analysis found inverse curvilinear dose-response for leisure-time PA and incident depression: overall RR=0.77 (95% CI=0.68, 0.86; I²=69%); light RR=0.73 (95% CI=0.64, 0.82; I²=43%), moderate RR=0.83 (95% CI=0.78, 0.87; I²=46%), highest RR=0.93 (95% CI=0.86, 0.99; I²=79%). Continuous dose-response: <25 MET-h/week—RR decreases 3% per 5 MET-h/week (RR=0.97, 95% CI=0.95, 0.98); >25 MET-h/week—no further reduction (RR=1.04, 95% CI=1.02, 1.05). Guo et al. (2022) found a nonlinear leisure-time PA–depression relationship. Zimmermann et al. (2020) reported mixed findings for anxiety (some protective effects for specific disorders). Dishman et al. (2021) reported reduced odds for incident depression/subclinical symptoms (OR=0.69; adjusted OR=0.79) but mixed outcome definitions limited extraction. - Causality assessment (Bradford Hill): Criteria supported temporality, consistency, biological plausibility, coherence, and some evidence of dose-response; specificity not met but of limited relevance. Overall grade: probable causal relationship—higher PA probably lowers risk of incident depression and anxiety.

Findings consistently show that higher physical activity is associated with lower incidence of depression and anxiety across ages and regions. Applying Bradford Hill criteria, the associations likely reflect a causal relationship, supporting PA as a preventive strategy for mental ill health. Although evidence is primarily observational, varied measurement of PA and outcomes, and potential publication bias, adjustments and subgroup analyses suggest the associations are robust. The public health implications are substantial given the global burden of depression and anxiety and high prevalence of insufficient PA. Integrating PA promotion (including active transport) into prevention policies could yield significant health and equity benefits, complementing reductions in other chronic diseases. However, more rigorous studies (e.g., randomized trials where feasible, longitudinal cohorts with objective PA measurement and repeated exposure assessments) are needed to refine dose-response estimates, understand subgroup effects (sex, age, socioeconomic status, ethnicity), and strengthen causal inference.

Evidence from systematic reviews and meta-analyses of prospective cohorts indicates that physical activity is probably causally associated with reduced incidence of depression and anxiety. While further high-quality research, including randomized controlled trials and cohorts with objective, repeated PA measures, is warranted, current findings support inclusion of PA in strategies to prevent mental ill health. Enhancing population PA levels could yield meaningful mental health benefits alongside well-established physical health gains.

- Protocol not registered; restricted to peer-reviewed, English-language reviews, potentially introducing publication bias. - Overlap of primary studies across included reviews; most primary cohorts from high-income settings, limiting generalizability to low- and middle-income countries. - Predominant reliance on self-reported PA (exposure misclassification, regression dilution, recall bias); many studies measured PA only at baseline. - Heterogeneous PA definitions and instruments (intensity, frequency, volume), limiting precise dose-response estimation. - Variation in outcome measurement (screens vs clinical diagnosis), though many used validated instruments. - Potential residual confounding (e.g., genetics, family history, obesity, diet, tobacco use, comorbidities) not consistently addressed. - Limited number of studies for incident anxiety; some analyses showed moderate heterogeneity and publication bias.