Abstract Whether the systemic manifestations observed in Chronic Obstructive Pulmonary Disease (COPD) are ascribable to lung dysfunction or direct effects of smoking is in debate. Structural Equations Modeling (SEM), a causal-oriented statistical approach, could help unraveling the pathways involved, by enabling estimation of direct and indirect associations between variables. The objectives of the study was to investigate the relative impact of smoking and COPD on systemic manifestations, inflammation and telomere length. In 292 individuals (103 women; 97 smokers with COPD, 96 smokers without COPD, 99 non-smokers), we used SEM to explore the pathways between smoking (pack-years), lung disease (FEV 1 , K CO ), and the following parameters: arterial stiffness (aortic pulse wave velocity, PWV), bone mineral density (BMD), appendicular skeletal muscle mass (ASMM), grip strength, insulin resistance (HOMA-IR), creatinine clearance (eGFR), blood leukocyte telomere length and inflammatory markers (Luminex assay). All models were adjusted on age and gender. Latent variables were created for systemic inflammation (inflammatory markers) and musculoskeletal parameters (ASMM, grip strength, BMD). SEM showed that most effects of smoking were indirectly mediated by lung dysfunction: e.g. via FEV 1 on musculoskeletal factor, eGFR, HOMA-IR, PWV, telomere length, CRP, white blood cells count (WBC) and inflammation factor, and via K CO on musculoskeletal factor, eGFR and PWV. Direct effects of smoking were limited to CRP and WBC. Models had excellent fit. In conclusion, SEM highlighted the major role of COPD in the occurrence of systemic manifestations while smoking effects were mostly mediated by lung function.