Background: Brain natriuretic peptide (BNP) predicts exercise performance and exercise training may modulate BNP and its N-terminal portion (NT-pro-BNP), we therefore conducted an individual patient analysis of exercise training effects on BNP and NT-pro-BNP. Aims: To use an individual patient meta-analysis to relate changes in BNP, NT-pro-BNP, and peak VO₂; to link these changes to volume parameters of exercise training programmes (intensity etc.); and to identify patient characteristics likely to lead to greater improvements in BNP, NT-pro-BNP, and peak VO₂. Design: Individual patient meta-analysis. Methods: A systematic search was conducted of Medline (Ovid), Embase.com, Cochrane Central Register of Controlled Trials, and CINAHL (until July 2008) to identify randomized controlled trials of aerobic and/or resistance exercise training in systolic heart failure patients measuring BNP and/or NT-pro-BNP. Primary outcome measures were change in BNP, NT-pro-BNP, and peak VO₂. Subanalyses were conducted to identify (1) patient groups that benefit most and (2) exercise programme parameters enhancing favourable changes in primary outcome measures. Results: Ten randomized controlled studies measuring BNP or NT-pro-BNP met eligibility criteria, authors provided individual patient data for 565 patients (313 exercise and 252 controls). Exercise training had favourable effects on BNP (-28.3%, p < 0.0001), NT-pro-BNP (-37.4%, p = < 0.0001), and peak VO₂ (17.8%, p < 0.0001). The analysis showed a significant change in primary outcome measures; moreover, change in BNP (r = -0.31, p < 0.0001) and NT-pro-BNP (r = -0.22, p < 0.0001) were correlated with peak VO₂ change. Conclusion: Exercise training has favourable effects on BNP, NT-pro-BNP, and peak VO₂ in heart failure patients and BNP/NT-pro-BNP changes were correlated with peak VO₂ changes.