Effects of exercise therapy on anxiety and depression in patients with COVID-19: a systematic review and meta-analysis

SARS-CoV-2 caused a global COVID-19 pandemic with substantial morbidity, mortality, and ongoing variants (e.g., Omicron) challenging treatment. Beyond physical illness, COVID-19 has profound psychological impacts; patients show markedly elevated risks of anxiety and depression, with prevalence estimates over 40%. Although pharmacologic treatments exist, many patients do not respond adequately, cannot tolerate side effects, or experience withdrawal upon discontinuation, prompting interest in non-pharmacologic therapies. Exercise has demonstrated benefits for mental health, including in randomized trials among those with anxiety and depression, and may serve as a flexible and accessible therapeutic option for COVID-19 patients. This systematic review and meta-analysis aimed to evaluate whether exercise therapy improves anxiety and depression among patients with COVID-19 and to inform management of post-COVID-19 mental health challenges.

Prior research indicates elevated anxiety, depression, and sleep disturbances among COVID-19 patients. Non-pharmacologic therapies such as exercise, yoga, and meditation have shown benefits for mood disorders, sometimes comparable to psychotherapy or pharmacotherapy, with added advantages of low cost and minimal side effects. Observational and interventional studies suggest physical activity reduces anxiety and depressive symptoms, with higher activity levels linked to lower symptom burden. Mechanistic literature links depression and anxiety to inflammation, oxidative stress, HPA axis dysregulation, and gut microbiota imbalance; exercise is reported to reduce oxidative damage and inflammatory markers, modulate HPA axis activity, and beneficially reshape gut microbiota, providing plausible pathways for mental health improvement.

Design and registration: Systematic review and meta-analysis conducted per PRISMA guidelines; protocol registered in PROSPERO (CRD42023406439). Data sources and search: PubMed, EMBASE, Web of Science, and Cochrane Library searched from inception to April 30, 2023, using terms for COVID-19, depression, anxiety, exercise therapy, and RCTs, combining free-text and MeSH. Detailed strategies in Supplementary File S1. Eligibility criteria: Inclusion—(1) Population: COVID-19 patients with symptoms of anxiety and/or depression; (2) Study type: randomized controlled trials; (3) Primary outcomes: degree of anxiety and depression; (4) Secondary outcomes: PHQ-9 and sleep quality; (5) Published within the search period. Exclusion—reviews, protocols, meta-analyses, non-RCTs, RCTs without outcome data, non-COVID-19 populations, studies without anxiety/depression, non-exercise interventions, letters/comments/abstracts/editorials/dissertations. Interventions and comparators: Experimental groups received exercise therapy (e.g., respiratory muscle training, stretching/home exercise, yoga/meditation, pulmonary rehabilitation with Baduanjin/five-tone breathing, physical activity programs, Baduanjin Qigong with music therapy). Control groups received routine care or symptomatic treatment without exercise therapy. Outcomes: Anxiety measured by SAS, HAM-A, SAI, GAD-7, HADS; depression by HADS, SDS, BDI; secondary outcomes included PHQ-9 and sleep quality (PSQI, ISI-K). Study selection: Two reviewers independently screened titles/abstracts and full texts, with EndNote X9 used for deduplication; disagreements resolved by a third reviewer. Data extraction: Two reviewers independently extracted study characteristics (year, author, sample size, age, interventions, comparators, treatment duration, outcomes). Discrepancies resolved via third reviewer. Risk of bias: Two researchers assessed risk of bias using the Cochrane RoB tool (seven domains), categorizing as low, unclear, or high risk. Statistical analysis: RevMan 5.4 used. Continuous outcomes synthesized using mean difference (MD) or standardized mean difference (SMD) with 95% CIs, selecting SMD when different scales were used. Random-effects model applied. Heterogeneity assessed with I² (I² <25% none; 25–50% moderate; >50% substantial). Subgroup analyses conducted when I² ≥25% (e.g., intervention duration <1 month vs ≥1 month; exercise modality).

- Study pool: 6 RCTs; total N = 461 participants with COVID-19. - Anxiety: Exercise therapy significantly reduced anxiety vs control (SMD = −0.76; 95% CI: −0.96 to −0.55; p < 0.00001; low heterogeneity, I² ≈ 14%). - Depression: Exercise therapy significantly improved depressive symptoms (SMD = −0.39; 95% CI: −0.70 to −0.09; p = 0.01; I² = 43%). - PHQ-9: In 2 studies, exercise improved PHQ-9 scores (MD = −1.82; 95% CI: −2.93 to −0.71; p = 0.001; I² = 0%). - Sleep quality: In 3 studies, exercise improved sleep quality (SMD = −0.73; 95% CI: −1.32 to −0.14; p = 0.01; I² = 81%). - Subgroups (depression outcome): • Duration: ≥1 month interventions showed significant improvement (pooled SMD = −0.60; 95% CI: −1.10 to −0.11; p = 0.02), while <1 month did not show a significant effect (SMD = −0.21; 95% CI: −0.51 to 0.09). • Modality: Breathing-focused intervention alone was not significant (single study SMD = −0.20; p = 0.40), whereas exercise-focused interventions showed benefit (pooled SMD = −0.46; 95% CI: −0.85 to −0.07; p = 0.02). - Risk of bias: One quasi-randomized study (Sharma) rated high risk for randomization; allocation concealment and blinding of outcome assessment were generally unclear; outcome data largely complete; selective reporting risk often unclear. - Overall: Exercise therapy was associated with clinically meaningful reductions in anxiety and depression and improvements in sleep among COVID-19 patients.

The findings directly address the question of whether exercise therapy benefits mental health in COVID-19 patients, demonstrating consistent reductions in anxiety and depressive symptoms and improvements in sleep quality. These results support exercise as a feasible, non-pharmacologic strategy during a time when effective antivirals remain limited and mental health burdens have surged. Exercise may confer benefits through biological mechanisms implicated in mood disorders, including reductions in oxidative stress and systemic inflammation, modulation of the HPA axis, and favorable shifts in gut microbiota composition and gut–brain axis function. The subgroup results suggest treatment duration matters—programs lasting at least one month may be needed to realize depression benefits—and that exercise-focused modalities are particularly effective compared with breathing-only approaches. From a public health standpoint, integrating structured exercise into COVID-19 care pathways could offer low-cost, acceptable, and scalable support for mental health recovery, potentially reducing the burden on healthcare systems and improving quality of life. Nevertheless, heterogeneity in interventions and multimodal components in several trials warrant cautious interpretation of effect attribution solely to exercise.

Across six RCTs involving 461 participants, exercise therapy significantly improved anxiety and depression in COVID-19 patients and also enhanced PHQ-9 scores and sleep quality. Longer intervention duration (≥1 month) and exercise-focused programs may yield greater benefits for depressive symptoms. Exercise therapy represents a practical, low-cost adjunct for managing post-COVID-19 mental health sequelae. Future research should include larger, high-quality, adequately blinded RCTs with standardized exercise protocols, diverse age groups, and designs that isolate the effects of exercise alone to clarify optimal dosing, modality, and long-term outcomes.

- Limited number of included studies (n=6) and small to moderate sample sizes. - Use of varied anxiety and depression scales necessitated SMD pooling, complicating direct clinical interpretation. - Age of participants predominantly over 40 years, limiting generalizability to younger populations. - Five of six trials were multimodal interventions emphasizing exercise, making it difficult to isolate the pure effect of exercise therapy alone. - Several risk-of-bias domains (e.g., allocation concealment, blinding of outcome assessment) were unclear; one quasi-randomized study presented high risk for randomization. - Heterogeneity was moderate for some outcomes (e.g., sleep quality), reflecting variability in interventions and populations.