Abstract
Background: Brain-derived neurotrophic factor (BDNF) is a key regulator of neuroplasticity, cognitive function, and mental health. Exercise is proposed as a non-pharmacological strategy to enhance BDNF; however, findings across neurological and non-neurological disorders remain inconsistent, and the influence of exercise type or dose-related parameters remains unclear. Objective: This meta-analysis evaluated the effects of exercise interventions on peripheral BDNF levels in individuals with neurological (e.g., multiple sclerosis, cognitive impairment, schizophrenia, depression) and non-neurological (e.g., obesity, type 2 diabetes, cancer) disorders, and examined whether outcomes varied by disease category, exercise modality, or dose. Methods: A systematic search of Web of Science, PubMed, ScienceDirect, Scopus, and Cochrane was conducted up to 1 October 2025. Eligible randomized controlled trials (RCTs) and the quality of evidence were assessed using the PEDro scale and the GRADE approach, respectively. Random-effects models were applied, with subgroup analyses (neurological vs. non-neurological; exercise type; duration and assay type), meta-regressions (duration, frequency, session length), and publication bias tests (funnel plot, Begg’s test, Egger’s regression, and trim-and-fill). Results: Nineteen RCTs, including 850 participants, were analyzed. According to low-quality evidence, exercise significantly increased peripheral BDNF (SMD = 1.03, 95% CI: [0.56–1.49, p < 0.0001). Effects did not differ significantly between neurological (SMD = 0.91, 95% CI: 0.31–1.50) and non-neurological (SMD = 1.23, 95% CI: 0.47–1.99) conditions (Q (1) = 0.44, p = 0.51). Subgroup analyses revealed significant improvements for resistance exercise (SMD = 1.57, 95% CI: 0.91–2.23), followed by aerobic (SMD = 1.44, 95% CI: 0.36–2.52) and combined exercise (SMD = 0.55, 95% CI: 0.21–0.89). Meta-regressions showed no moderating effects of duration (β = 0.0101, p = 0.834), weekly frequency (β = 0.1464, p = 0.648), minutes per session (β = −0.0124, p = 0.233) or total weekly minutes (β = 0.0005, p = 0.919) apart from age and baseline BDNF level factors (β = 0.0348, p = 0.020; β = −0.035, p = 0.0258). Publication bias tests indicated minimal publication bias, with adjusted effects remaining robust. Conclusions: Exercise interventions have been shown to increase peripheral BDNF significantly across diverse clinical populations. In particular, resistance and aerobic exercise protocols accounted for the exploratory component, whereas simple dose-related factors did not explain the variability. These findings are consistent with the biological plausibility of exercise-induced neuroplasticity and underscore the need for larger, pre-registered RCTs with harmonized biomarker protocols to strengthen clinical translation. However, the certainty of evidence is limited by small sample sizes and the frequent lack of blinding of participants and assessors across included trials.